code
stringlengths 82
54.1k
| code_codestyle
int64 0
699
| style_context
stringlengths 111
35.6k
| style_context_codestyle
int64 0
699
| label
int64 0
1
|
|---|---|---|---|---|
import inspect
import unittest
import warnings
from math import ceil, floor
from transformers import LevitConfig
from transformers.file_utils import cached_property, is_torch_available, is_vision_available
from transformers.models.auto import get_values
from transformers.testing_utils import require_torch, require_vision, slow, torch_device
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from transformers import (
MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
MODEL_MAPPING,
LevitForImageClassification,
LevitForImageClassificationWithTeacher,
LevitModel,
)
from transformers.models.levit.modeling_levit import LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
from PIL import Image
from transformers import LevitImageProcessor
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Union[str, Any] ):
a__ : Union[str, Any] = self.config_class(**self.inputs_dict )
self.parent.assertTrue(hasattr(lowerCamelCase__ , "hidden_sizes" ) )
self.parent.assertTrue(hasattr(lowerCamelCase__ , "num_attention_heads" ) )
class A__ :
"""simple docstring"""
def __init__( self : int , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[Any]=13 , lowerCamelCase__ : Dict=64 , lowerCamelCase__ : List[str]=3 , lowerCamelCase__ : List[str]=3 , lowerCamelCase__ : Union[str, Any]=2 , lowerCamelCase__ : int=1 , lowerCamelCase__ : Union[str, Any]=16 , lowerCamelCase__ : List[str]=[128, 256, 384] , lowerCamelCase__ : Optional[int]=[4, 6, 8] , lowerCamelCase__ : Optional[int]=[2, 3, 4] , lowerCamelCase__ : Tuple=[16, 16, 16] , lowerCamelCase__ : Optional[int]=0 , lowerCamelCase__ : Optional[Any]=[2, 2, 2] , lowerCamelCase__ : int=[2, 2, 2] , lowerCamelCase__ : Optional[int]=0.02 , lowerCamelCase__ : Union[str, Any]=True , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : Optional[Any]=2 , ):
a__ : Union[str, Any] = parent
a__ : List[Any] = batch_size
a__ : Dict = image_size
a__ : Any = num_channels
a__ : Union[str, Any] = kernel_size
a__ : str = stride
a__ : List[str] = padding
a__ : Tuple = hidden_sizes
a__ : Any = num_attention_heads
a__ : List[str] = depths
a__ : Any = key_dim
a__ : Tuple = drop_path_rate
a__ : Union[str, Any] = patch_size
a__ : Optional[int] = attention_ratio
a__ : int = mlp_ratio
a__ : Union[str, Any] = initializer_range
a__ : List[Any] = [
["Subsample", key_dim[0], hidden_sizes[0] // key_dim[0], 4, 2, 2],
["Subsample", key_dim[0], hidden_sizes[1] // key_dim[0], 4, 2, 2],
]
a__ : Tuple = is_training
a__ : List[str] = use_labels
a__ : List[Any] = num_labels
a__ : List[Any] = initializer_range
def _UpperCamelCase( self : Dict ):
a__ : str = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
a__ : Any = None
if self.use_labels:
a__ : Tuple = ids_tensor([self.batch_size] , self.num_labels )
a__ : Union[str, Any] = self.get_config()
return config, pixel_values, labels
def _UpperCamelCase( self : Union[str, Any] ):
return LevitConfig(
image_size=self.image_size , num_channels=self.num_channels , kernel_size=self.kernel_size , stride=self.stride , padding=self.padding , patch_size=self.patch_size , hidden_sizes=self.hidden_sizes , num_attention_heads=self.num_attention_heads , depths=self.depths , key_dim=self.key_dim , drop_path_rate=self.drop_path_rate , mlp_ratio=self.mlp_ratio , attention_ratio=self.attention_ratio , initializer_range=self.initializer_range , down_ops=self.down_ops , )
def _UpperCamelCase( self : Any , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : List[str] ):
a__ : List[Any] = LevitModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Dict = model(lowerCamelCase__ )
a__ : Optional[Any] = (self.image_size, self.image_size)
a__, a__ : Optional[Any] = image_size[0], image_size[1]
for _ in range(4 ):
a__ : List[str] = floor(((height + 2 * self.padding - self.kernel_size) / self.stride) + 1 )
a__ : Union[str, Any] = floor(((width + 2 * self.padding - self.kernel_size) / self.stride) + 1 )
self.parent.assertEqual(
result.last_hidden_state.shape , (self.batch_size, ceil(height / 4 ) * ceil(width / 4 ), self.hidden_sizes[-1]) , )
def _UpperCamelCase( self : Any , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : List[str] ):
a__ : str = self.num_labels
a__ : Tuple = LevitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : int = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) )
def _UpperCamelCase( self : str ):
a__ : List[str] = self.prepare_config_and_inputs()
a__, a__, a__ : str = config_and_inputs
a__ : Union[str, Any] = {"pixel_values": pixel_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(LevitModel, LevitForImageClassification, LevitForImageClassificationWithTeacher)
if is_torch_available()
else ()
)
_lowercase = (
{
'feature-extraction': LevitModel,
'image-classification': (LevitForImageClassification, LevitForImageClassificationWithTeacher),
}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Optional[int] ):
a__ : Optional[Any] = LevitModelTester(self )
a__ : Union[str, Any] = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : Optional[int] ):
self.create_and_test_config_common_properties()
self.config_tester.create_and_test_config_to_json_string()
self.config_tester.create_and_test_config_to_json_file()
self.config_tester.create_and_test_config_from_and_save_pretrained()
self.config_tester.create_and_test_config_with_num_labels()
self.config_tester.check_config_can_be_init_without_params()
self.config_tester.check_config_arguments_init()
def _UpperCamelCase( self : Dict ):
return
@unittest.skip(reason="Levit does not use inputs_embeds" )
def _UpperCamelCase( self : Optional[int] ):
pass
@unittest.skip(reason="Levit does not support input and output embeddings" )
def _UpperCamelCase( self : Union[str, Any] ):
pass
@unittest.skip(reason="Levit does not output attentions" )
def _UpperCamelCase( self : Optional[int] ):
pass
def _UpperCamelCase( self : str ):
a__, a__ : int = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Dict = model_class(lowerCamelCase__ )
a__ : int = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Dict = [*signature.parameters.keys()]
a__ : Union[str, Any] = ["pixel_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : Dict ):
def check_hidden_states_output(lowerCamelCase__ : List[Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[int] ):
a__ : List[str] = model_class(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
with torch.no_grad():
a__ : Dict = model(**self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ ) )
a__ : Optional[int] = outputs.hidden_states
a__ : List[Any] = len(self.model_tester.depths ) + 1
self.assertEqual(len(lowerCamelCase__ ) , lowerCamelCase__ )
a__ : int = (self.model_tester.image_size, self.model_tester.image_size)
a__, a__ : List[Any] = image_size[0], image_size[1]
for _ in range(4 ):
a__ : str = floor(
(
(height + 2 * self.model_tester.padding - self.model_tester.kernel_size)
/ self.model_tester.stride
)
+ 1 )
a__ : Union[str, Any] = floor(
(
(width + 2 * self.model_tester.padding - self.model_tester.kernel_size)
/ self.model_tester.stride
)
+ 1 )
# verify the first hidden states (first block)
self.assertListEqual(
list(hidden_states[0].shape[-2:] ) , [
height * width,
self.model_tester.hidden_sizes[0],
] , )
a__, a__ : Any = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Union[str, Any] = True
check_hidden_states_output(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
# check that output_hidden_states also work using config
del inputs_dict["output_hidden_states"]
a__ : str = True
check_hidden_states_output(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
@unittest.skip("Will be fixed soon by reducing the size of the model used for common tests." )
def _UpperCamelCase( self : Optional[int] ):
pass
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : List[str]=False ):
a__ : Union[str, Any] = super()._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
if return_labels:
if model_class.__name__ == "LevitForImageClassificationWithTeacher":
del inputs_dict["labels"]
return inputs_dict
def _UpperCamelCase( self : List[str] ):
a__ : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
def _UpperCamelCase( self : str ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
if not self.model_tester.is_training:
return
a__, a__ : Optional[int] = self.model_tester.prepare_config_and_inputs_for_common()
a__ : Tuple = True
for model_class in self.all_model_classes:
# LevitForImageClassificationWithTeacher supports inference-only
if (
model_class in get_values(lowerCamelCase__ )
or model_class.__name__ == "LevitForImageClassificationWithTeacher"
):
continue
a__ : List[str] = model_class(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.train()
a__ : int = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : Optional[int] = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : str ):
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
if not self.model_tester.is_training:
return
a__ : Tuple = False
a__ : Any = True
for model_class in self.all_model_classes:
if model_class in get_values(lowerCamelCase__ ) or not model_class.supports_gradient_checkpointing:
continue
# LevitForImageClassificationWithTeacher supports inference-only
if model_class.__name__ == "LevitForImageClassificationWithTeacher":
continue
a__ : Any = model_class(lowerCamelCase__ )
model.gradient_checkpointing_enable()
model.to(lowerCamelCase__ )
model.train()
a__ : List[str] = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : Tuple = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : Any ):
a__, a__ : List[Any] = self.model_tester.prepare_config_and_inputs_for_common()
a__ : Union[str, Any] = [
{"title": "multi_label_classification", "num_labels": 2, "dtype": torch.float},
{"title": "single_label_classification", "num_labels": 1, "dtype": torch.long},
{"title": "regression", "num_labels": 1, "dtype": torch.float},
]
for model_class in self.all_model_classes:
if (
model_class
not in [
*get_values(lowerCamelCase__ ),
]
or model_class.__name__ == "LevitForImageClassificationWithTeacher"
):
continue
for problem_type in problem_types:
with self.subTest(msg=f'''Testing {model_class} with {problem_type['title']}''' ):
a__ : Optional[Any] = problem_type["title"]
a__ : Any = problem_type["num_labels"]
a__ : Tuple = model_class(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.train()
a__ : Optional[Any] = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
if problem_type["num_labels"] > 1:
a__ : Tuple = inputs["labels"].unsqueeze(1 ).repeat(1 , problem_type["num_labels"] )
a__ : List[Any] = inputs["labels"].to(problem_type["dtype"] )
# This tests that we do not trigger the warning form PyTorch "Using a target size that is different
# to the input size. This will likely lead to incorrect results due to broadcasting. Please ensure
# they have the same size." which is a symptom something in wrong for the regression problem.
# See https://github.com/huggingface/transformers/issues/11780
with warnings.catch_warnings(record=lowerCamelCase__ ) as warning_list:
a__ : Union[str, Any] = model(**lowerCamelCase__ ).loss
for w in warning_list:
if "Using a target size that is different to the input size" in str(w.message ):
raise ValueError(
f'''Something is going wrong in the regression problem: intercepted {w.message}''' )
loss.backward()
@slow
def _UpperCamelCase( self : int ):
for model_name in LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Tuple = LevitModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> List[Any]:
a__ : Any = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
return image
@require_torch
@require_vision
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : Optional[Any] ):
return LevitImageProcessor.from_pretrained(LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST[0] )
@slow
def _UpperCamelCase( self : List[Any] ):
a__ : List[str] = LevitForImageClassificationWithTeacher.from_pretrained(LEVIT_PRETRAINED_MODEL_ARCHIVE_LIST[0] ).to(
lowerCamelCase__ )
a__ : int = self.default_image_processor
a__ : str = prepare_img()
a__ : str = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Dict = model(**lowerCamelCase__ )
# verify the logits
a__ : Tuple = torch.Size((1, 1_000) )
self.assertEqual(outputs.logits.shape , lowerCamelCase__ )
a__ : Dict = torch.tensor([1.0448, -0.3745, -1.8317] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(outputs.logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
| 37 |
from statistics import mean, stdev
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : List[str] = min(__a )
a__ : str = max(__a )
# normalize data
return [round((x - x_min) / (x_max - x_min) , __a ) for x in data]
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : str = mean(__a )
a__ : List[str] = stdev(__a )
# standardize data
return [round((x - mu) / (sigma) , __a ) for x in data]
| 37 | 1 |
from __future__ import annotations
import unittest
import numpy as np
from transformers import LayoutLMConfig, is_tf_available
from transformers.testing_utils import require_tf, slow
from ...test_configuration_common import ConfigTester
from ...test_modeling_tf_common import TFModelTesterMixin, ids_tensor, random_attention_mask
from ...test_pipeline_mixin import PipelineTesterMixin
if is_tf_available():
import tensorflow as tf
from transformers.models.layoutlm.modeling_tf_layoutlm import (
TF_LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST,
TFLayoutLMForMaskedLM,
TFLayoutLMForQuestionAnswering,
TFLayoutLMForSequenceClassification,
TFLayoutLMForTokenClassification,
TFLayoutLMModel,
)
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : Tuple=13 , lowerCamelCase__ : List[Any]=7 , lowerCamelCase__ : Any=True , lowerCamelCase__ : Any=True , lowerCamelCase__ : Dict=True , lowerCamelCase__ : Any=True , lowerCamelCase__ : Optional[Any]=99 , lowerCamelCase__ : Tuple=32 , lowerCamelCase__ : Union[str, Any]=2 , lowerCamelCase__ : Any=4 , lowerCamelCase__ : int=37 , lowerCamelCase__ : Tuple="gelu" , lowerCamelCase__ : int=0.1 , lowerCamelCase__ : str=0.1 , lowerCamelCase__ : Optional[int]=512 , lowerCamelCase__ : str=16 , lowerCamelCase__ : List[Any]=2 , lowerCamelCase__ : int=0.02 , lowerCamelCase__ : List[Any]=3 , lowerCamelCase__ : List[Any]=4 , lowerCamelCase__ : Optional[Any]=None , lowerCamelCase__ : int=1_000 , ):
a__ : Union[str, Any] = parent
a__ : Optional[Any] = batch_size
a__ : Optional[int] = seq_length
a__ : Any = is_training
a__ : Tuple = use_input_mask
a__ : str = use_token_type_ids
a__ : int = use_labels
a__ : int = vocab_size
a__ : int = hidden_size
a__ : List[str] = num_hidden_layers
a__ : Tuple = num_attention_heads
a__ : Any = intermediate_size
a__ : Tuple = hidden_act
a__ : List[Any] = hidden_dropout_prob
a__ : List[Any] = attention_probs_dropout_prob
a__ : Dict = max_position_embeddings
a__ : int = type_vocab_size
a__ : str = type_sequence_label_size
a__ : List[str] = initializer_range
a__ : List[str] = num_labels
a__ : Any = num_choices
a__ : Union[str, Any] = scope
a__ : Dict = range_bbox
def _UpperCamelCase( self : Optional[Any] ):
a__ : int = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
# convert bbox to numpy since TF does not support item assignment
a__ : Any = ids_tensor([self.batch_size, self.seq_length, 4] , self.range_bbox ).numpy()
# Ensure that bbox is legal
for i in range(bbox.shape[0] ):
for j in range(bbox.shape[1] ):
if bbox[i, j, 3] < bbox[i, j, 1]:
a__ : int = bbox[i, j, 3]
a__ : str = bbox[i, j, 1]
a__ : int = t
if bbox[i, j, 2] < bbox[i, j, 0]:
a__ : Union[str, Any] = bbox[i, j, 2]
a__ : int = bbox[i, j, 0]
a__ : Dict = t
a__ : List[str] = tf.convert_to_tensor(lowerCamelCase__ )
a__ : Union[str, Any] = None
if self.use_input_mask:
a__ : List[Any] = random_attention_mask([self.batch_size, self.seq_length] )
a__ : List[Any] = None
if self.use_token_type_ids:
a__ : Any = ids_tensor([self.batch_size, self.seq_length] , self.type_vocab_size )
a__ : str = None
a__ : str = None
a__ : int = None
if self.use_labels:
a__ : Optional[int] = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : int = ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
a__ : str = ids_tensor([self.batch_size] , self.num_choices )
a__ : Tuple = LayoutLMConfig(
vocab_size=self.vocab_size , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , max_position_embeddings=self.max_position_embeddings , type_vocab_size=self.type_vocab_size , initializer_range=self.initializer_range , )
return config, input_ids, bbox, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Dict , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : int , lowerCamelCase__ : List[Any] ):
a__ : Any = TFLayoutLMModel(config=lowerCamelCase__ )
a__ : Any = model(lowerCamelCase__ , lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ )
a__ : List[Any] = model(lowerCamelCase__ , lowerCamelCase__ , token_type_ids=lowerCamelCase__ )
a__ : Tuple = model(lowerCamelCase__ , lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
self.parent.assertEqual(result.pooler_output.shape , (self.batch_size, self.hidden_size) )
def _UpperCamelCase( self : int , lowerCamelCase__ : str , lowerCamelCase__ : Dict , lowerCamelCase__ : List[str] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : List[str] , lowerCamelCase__ : int , lowerCamelCase__ : int , lowerCamelCase__ : str ):
a__ : int = TFLayoutLMForMaskedLM(config=lowerCamelCase__ )
a__ : Union[str, Any] = model(lowerCamelCase__ , lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Any , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Dict , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Dict , lowerCamelCase__ : Tuple ):
a__ : Any = self.num_labels
a__ : Optional[Any] = TFLayoutLMForSequenceClassification(config=lowerCamelCase__ )
a__ : int = model(lowerCamelCase__ , lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_labels) )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Any , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : str , lowerCamelCase__ : Dict ):
a__ : Any = self.num_labels
a__ : Any = TFLayoutLMForTokenClassification(config=lowerCamelCase__ )
a__ : int = model(lowerCamelCase__ , lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Dict , lowerCamelCase__ : Any , lowerCamelCase__ : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : List[str] ):
a__ : Optional[int] = TFLayoutLMForQuestionAnswering(config=lowerCamelCase__ )
a__ : List[str] = model(lowerCamelCase__ , lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ )
self.parent.assertEqual(result.start_logits.shape , (self.batch_size, self.seq_length) )
self.parent.assertEqual(result.end_logits.shape , (self.batch_size, self.seq_length) )
def _UpperCamelCase( self : Dict ):
a__ : Union[str, Any] = self.prepare_config_and_inputs()
(
(
a__
), (
a__
), (
a__
), (
a__
), (
a__
), (
a__
), (
a__
), (
a__
),
) : List[Any] = config_and_inputs
a__ : Optional[int] = {
"input_ids": input_ids,
"bbox": bbox,
"token_type_ids": token_type_ids,
"attention_mask": input_mask,
}
return config, inputs_dict
@require_tf
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(
TFLayoutLMModel,
TFLayoutLMForMaskedLM,
TFLayoutLMForTokenClassification,
TFLayoutLMForSequenceClassification,
TFLayoutLMForQuestionAnswering,
)
if is_tf_available()
else ()
)
_lowercase = (
{
'feature-extraction': TFLayoutLMModel,
'fill-mask': TFLayoutLMForMaskedLM,
'text-classification': TFLayoutLMForSequenceClassification,
'token-classification': TFLayoutLMForTokenClassification,
'zero-shot': TFLayoutLMForSequenceClassification,
}
if is_tf_available()
else {}
)
_lowercase = False
_lowercase = True
_lowercase = 1_0
def _UpperCamelCase( self : List[str] ):
a__ : Optional[int] = TFLayoutLMModelTester(self )
a__ : Union[str, Any] = ConfigTester(self , config_class=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[str] ):
self.config_tester.run_common_tests()
def _UpperCamelCase( self : int ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : Tuple = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_lm(*lowerCamelCase__ )
def _UpperCamelCase( self : Any ):
a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_sequence_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : str ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_token_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : int ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_question_answering(*lowerCamelCase__ )
@slow
def _UpperCamelCase( self : Tuple ):
for model_name in TF_LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : str = TFLayoutLMModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
@unittest.skip("Onnx compliancy broke with TF 2.10" )
def _UpperCamelCase( self : Optional[int] ):
pass
def UpperCamelCase_ ( ) -> List[Any]:
# Here we prepare a batch of 2 sequences to test a LayoutLM forward pass on:
# fmt: off
a__ : Optional[Any] = tf.convert_to_tensor([[101,1_019,1_014,1_016,1_037,12_849,4_747,1_004,14_246,2_278,5_439,4_524,5_002,2_930,2_193,2_930,4_341,3_208,1_005,1_055,2_171,2_848,11_300,3_531,102],[101,4_070,4_034,7_020,1_024,3_058,1_015,1_013,2_861,1_013,6_070,19_274,2_772,6_205,27_814,16_147,16_147,4_343,2_047,10_283,10_969,14_389,1_012,2_338,102]] ) # noqa: E231
a__ : Optional[int] = tf.convert_to_tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],] ) # noqa: E231
a__ : Tuple = tf.convert_to_tensor([[[0,0,0,0],[423,237,440,251],[427,272,441,287],[419,115,437,129],[961,885,992,912],[256,38,330,58],[256,38,330,58],[336,42,353,57],[360,39,401,56],[360,39,401,56],[411,39,471,59],[479,41,528,59],[533,39,630,60],[67,113,134,131],[141,115,209,132],[68,149,133,166],[141,149,187,164],[195,148,287,165],[195,148,287,165],[195,148,287,165],[295,148,349,165],[441,149,492,166],[497,149,546,164],[64,201,125,218],[1_000,1_000,1_000,1_000]],[[0,0,0,0],[662,150,754,166],[665,199,742,211],[519,213,554,228],[519,213,554,228],[134,433,187,454],[130,467,204,480],[130,467,204,480],[130,467,204,480],[130,467,204,480],[130,467,204,480],[314,469,376,482],[504,684,582,706],[941,825,973,900],[941,825,973,900],[941,825,973,900],[941,825,973,900],[610,749,652,765],[130,659,168,672],[176,657,237,672],[238,657,312,672],[443,653,628,672],[443,653,628,672],[716,301,825,317],[1_000,1_000,1_000,1_000]]] ) # noqa: E231
a__ : Any = tf.convert_to_tensor([[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]] ) # noqa: E231
# these are sequence labels (i.e. at the token level)
a__ : List[str] = tf.convert_to_tensor([[-100,10,10,10,9,1,-100,7,7,-100,7,7,4,2,5,2,8,8,-100,-100,5,0,3,2,-100],[-100,12,12,12,-100,12,10,-100,-100,-100,-100,10,12,9,-100,-100,-100,10,10,10,9,12,-100,10,-100]] ) # noqa: E231
# fmt: on
return input_ids, attention_mask, bbox, token_type_ids, labels
@require_tf
class A__ ( unittest.TestCase ):
"""simple docstring"""
@slow
def _UpperCamelCase( self : int ):
a__ : Union[str, Any] = TFLayoutLMModel.from_pretrained("microsoft/layoutlm-base-uncased" )
a__, a__, a__, a__, a__ : Tuple = prepare_layoutlm_batch_inputs()
# forward pass
a__ : Optional[int] = model(input_ids=lowerCamelCase__ , bbox=lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ )
# test the sequence output on [0, :3, :3]
a__ : Dict = tf.convert_to_tensor(
[[0.1785, -0.1947, -0.0425], [-0.3254, -0.2807, 0.2553], [-0.5391, -0.3322, 0.3364]] , )
self.assertTrue(np.allclose(outputs.last_hidden_state[0, :3, :3] , lowerCamelCase__ , atol=1E-3 ) )
# test the pooled output on [1, :3]
a__ : Tuple = tf.convert_to_tensor([-0.6580, -0.0214, 0.8552] )
self.assertTrue(np.allclose(outputs.pooler_output[1, :3] , lowerCamelCase__ , atol=1E-3 ) )
@slow
def _UpperCamelCase( self : Optional[int] ):
# initialize model with randomly initialized sequence classification head
a__ : Tuple = TFLayoutLMForSequenceClassification.from_pretrained("microsoft/layoutlm-base-uncased" , num_labels=2 )
a__, a__, a__, a__, a__ : str = prepare_layoutlm_batch_inputs()
# forward pass
a__ : Optional[Any] = model(
input_ids=lowerCamelCase__ , bbox=lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ , labels=tf.convert_to_tensor([1, 1] ) , )
# test whether we get a loss as a scalar
a__ : Union[str, Any] = outputs.loss
a__ : Tuple = (2,)
self.assertEqual(loss.shape , lowerCamelCase__ )
# test the shape of the logits
a__ : Dict = outputs.logits
a__ : Any = (2, 2)
self.assertEqual(logits.shape , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : List[str] ):
# initialize model with randomly initialized token classification head
a__ : str = TFLayoutLMForTokenClassification.from_pretrained("microsoft/layoutlm-base-uncased" , num_labels=13 )
a__, a__, a__, a__, a__ : Optional[Any] = prepare_layoutlm_batch_inputs()
# forward pass
a__ : Optional[Any] = model(
input_ids=lowerCamelCase__ , bbox=lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ , labels=lowerCamelCase__ )
# test the shape of the logits
a__ : Any = outputs.logits
a__ : List[Any] = tf.convert_to_tensor((2, 25, 13) )
self.assertEqual(logits.shape , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
# initialize model with randomly initialized token classification head
a__ : Tuple = TFLayoutLMForQuestionAnswering.from_pretrained("microsoft/layoutlm-base-uncased" )
a__, a__, a__, a__, a__ : Union[str, Any] = prepare_layoutlm_batch_inputs()
# forward pass
a__ : List[str] = model(input_ids=lowerCamelCase__ , bbox=lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ )
# test the shape of the logits
a__ : Dict = tf.convert_to_tensor((2, 25) )
self.assertEqual(outputs.start_logits.shape , lowerCamelCase__ )
self.assertEqual(outputs.end_logits.shape , lowerCamelCase__ )
| 37 |
def UpperCamelCase_ ( __a = 50 ) -> int:
a__ : Tuple = [[0] * 3 for _ in range(length + 1 )]
for row_length in range(length + 1 ):
for tile_length in range(2 , 5 ):
for tile_start in range(row_length - tile_length + 1 ):
different_colour_ways_number[row_length][tile_length - 2] += (
different_colour_ways_number[row_length - tile_start - tile_length][
tile_length - 2
]
+ 1
)
return sum(different_colour_ways_number[length] )
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 | 1 |
from collections import defaultdict
from math import gcd
def UpperCamelCase_ ( __a = 1_500_000 ) -> int:
a__ : defaultdict = defaultdict(__a )
a__ : Optional[int] = 2
while 2 * euclid_m * (euclid_m + 1) <= limit:
for euclid_n in range((euclid_m % 2) + 1 , __a , 2 ):
if gcd(__a , __a ) > 1:
continue
a__ : Any = 2 * euclid_m * (euclid_m + euclid_n)
for perimeter in range(__a , limit + 1 , __a ):
frequencies[perimeter] += 1
euclid_m += 1
return sum(1 for frequency in frequencies.values() if frequency == 1 )
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 |
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[str] ):
a__ : str = name
a__ : Optional[int] = value
a__ : Dict = weight
def __repr__( self : Union[str, Any] ):
return f'''{self.__class__.__name__}({self.name}, {self.value}, {self.weight})'''
def _UpperCamelCase( self : Dict ):
return self.value
def _UpperCamelCase( self : Optional[Any] ):
return self.name
def _UpperCamelCase( self : Optional[Any] ):
return self.weight
def _UpperCamelCase( self : Optional[int] ):
return self.value / self.weight
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = []
for i in range(len(__a ) ):
menu.append(Things(name[i] , value[i] , weight[i] ) )
return menu
def UpperCamelCase_ ( __a , __a , __a ) -> Union[str, Any]:
a__ : List[str] = sorted(__a , key=__a , reverse=__a )
a__ : List[Any] = []
a__, a__ : Union[str, Any] = 0.0, 0.0
for i in range(len(__a ) ):
if (total_cost + items_copy[i].get_weight()) <= max_cost:
result.append(items_copy[i] )
total_cost += items_copy[i].get_weight()
total_value += items_copy[i].get_value()
return (result, total_value)
def UpperCamelCase_ ( ) -> Union[str, Any]:
pass
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 | 1 |
def UpperCamelCase_ ( __a = 1_000_000 ) -> int:
a__ : int = limit + 1
a__ : str = [0] * limit
for first_term in range(1 , __a ):
for n in range(__a , __a , __a ):
a__ : Optional[int] = first_term + n / first_term
if common_difference % 4: # d must be divisble by 4
continue
else:
common_difference /= 4
if (
first_term > common_difference
and first_term < 4 * common_difference
): # since x,y,z are positive integers
frequency[n] += 1 # so z>0 and a>d ,also 4d<a
a__ : Tuple = sum(1 for x in frequency[1:limit] if x == 10 )
return count
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 |
import multiprocessing
from typing import TYPE_CHECKING, Optional, Union
from .. import Dataset, Features, config
from ..formatting import query_table
from ..packaged_modules.sql.sql import Sql
from ..utils import logging
from .abc import AbstractDatasetInputStream
if TYPE_CHECKING:
import sqlitea
import sqlalchemy
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Dict , lowerCamelCase__ : Union[str, "sqlalchemy.sql.Selectable"] , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[Features] = None , lowerCamelCase__ : str = None , lowerCamelCase__ : bool = False , **lowerCamelCase__ : Optional[int] , ):
super().__init__(features=lowerCamelCase__ , cache_dir=lowerCamelCase__ , keep_in_memory=lowerCamelCase__ , **lowerCamelCase__ )
a__ : str = Sql(
cache_dir=lowerCamelCase__ , features=lowerCamelCase__ , sql=lowerCamelCase__ , con=lowerCamelCase__ , **lowerCamelCase__ , )
def _UpperCamelCase( self : Tuple ):
a__ : Optional[Any] = None
a__ : Dict = None
a__ : Union[str, Any] = None
a__ : Union[str, Any] = None
self.builder.download_and_prepare(
download_config=lowerCamelCase__ , download_mode=lowerCamelCase__ , verification_mode=lowerCamelCase__ , base_path=lowerCamelCase__ , )
# Build dataset for splits
a__ : List[str] = self.builder.as_dataset(
split="train" , verification_mode=lowerCamelCase__ , in_memory=self.keep_in_memory )
return dataset
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : Dataset , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[int] = None , **lowerCamelCase__ : Optional[Any] , ):
if num_proc is not None and num_proc <= 0:
raise ValueError(f'''num_proc {num_proc} must be an integer > 0.''' )
a__ : Any = dataset
a__ : str = name
a__ : Tuple = con
a__ : List[Any] = batch_size if batch_size else config.DEFAULT_MAX_BATCH_SIZE
a__ : Any = num_proc
a__ : Tuple = to_sql_kwargs
def _UpperCamelCase( self : List[Any] ):
a__ : Any = self.to_sql_kwargs.pop("sql" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("con" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("index" , lowerCamelCase__ )
a__ : int = self._write(index=lowerCamelCase__ , **self.to_sql_kwargs )
return written
def _UpperCamelCase( self : Any , lowerCamelCase__ : List[str] ):
a__, a__, a__ : Union[str, Any] = args
a__ : Any = {**to_sql_kwargs, "if_exists": "append"} if offset > 0 else to_sql_kwargs
a__ : Tuple = query_table(
table=self.dataset.data , key=slice(lowerCamelCase__ , offset + self.batch_size ) , indices=self.dataset._indices , )
a__ : str = batch.to_pandas()
a__ : List[Any] = df.to_sql(self.name , self.con , index=lowerCamelCase__ , **lowerCamelCase__ )
return num_rows or len(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : str = 0
if self.num_proc is None or self.num_proc == 1:
for offset in logging.tqdm(
range(0 , len(self.dataset ) , self.batch_size ) , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += self._batch_sql((offset, index, to_sql_kwargs) )
else:
a__, a__ : List[str] = len(self.dataset ), self.batch_size
with multiprocessing.Pool(self.num_proc ) as pool:
for num_rows in logging.tqdm(
pool.imap(
self._batch_sql , [(offset, index, to_sql_kwargs) for offset in range(0 , lowerCamelCase__ , lowerCamelCase__ )] , ) , total=(num_rows // batch_size) + 1 if num_rows % batch_size else num_rows // batch_size , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += num_rows
return written
| 37 | 1 |
from __future__ import annotations
from dataclasses import dataclass
@dataclass
class A__ :
"""simple docstring"""
_lowercase = 42
_lowercase = None
_lowercase = None
def UpperCamelCase_ ( __a ) -> bool:
# Validation
def is_valid_tree(__a ) -> bool:
if node is None:
return True
if not isinstance(__a , __a ):
return False
try:
float(node.data )
except (TypeError, ValueError):
return False
return is_valid_tree(node.left ) and is_valid_tree(node.right )
if not is_valid_tree(__a ):
raise ValueError(
"Each node should be type of TreeNode and data should be float." )
def is_binary_search_tree_recursive_check(
__a , __a , __a ) -> bool:
if node is None:
return True
return (
left_bound < node.data < right_bound
and is_binary_search_tree_recursive_check(node.left , __a , node.data )
and is_binary_search_tree_recursive_check(
node.right , node.data , __a )
)
return is_binary_search_tree_recursive_check(__a , -float("inf" ) , float("inf" ) )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 |
import math
from datetime import datetime, timedelta
def UpperCamelCase_ ( __a ) -> datetime:
a__ : Union[str, Any] = year % 19
a__ : List[str] = year % 4
a__ : str = year % 7
a__ : Any = math.floor(year / 100 )
a__ : List[str] = math.floor((13 + 8 * leap_day_inhibits) / 25 )
a__ : Optional[int] = leap_day_inhibits / 4
a__ : Union[str, Any] = (
15 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 30
a__ : Dict = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
a__ : Any = (19 * metonic_cycle + secular_moon_shift) % 30
# PHM -> Paschal Full Moon
a__ : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 29 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 19 )
elif days_to_add == 28 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 18 )
else:
return datetime(__a , 3 , 22 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
UpperCamelCase : Tuple = """will be""" if year > datetime.now().year else """was"""
print(f"""Easter in {year} {tense} {gauss_easter(year)}""")
| 37 | 1 |
import re
from filelock import FileLock
try:
import nltk
UpperCamelCase : Union[str, Any] = True
except (ImportError, ModuleNotFoundError):
UpperCamelCase : Optional[int] = False
if NLTK_AVAILABLE:
with FileLock(""".lock""") as lock:
nltk.download("""punkt""", quiet=True)
def UpperCamelCase_ ( __a ) -> str:
re.sub("<n>" , "" , __a ) # remove pegasus newline char
assert NLTK_AVAILABLE, "nltk must be installed to separate newlines between sentences. (pip install nltk)"
return "\n".join(nltk.sent_tokenize(__a ) )
| 37 |
import gc
import importlib.metadata
import tempfile
import unittest
from packaging import version
from transformers import (
AutoModel,
AutoModelForCausalLM,
AutoModelForSeqaSeqLM,
AutoModelForSequenceClassification,
AutoTokenizer,
BitsAndBytesConfig,
pipeline,
)
from transformers.testing_utils import (
is_torch_available,
require_accelerate,
require_bitsandbytes,
require_torch,
require_torch_gpu,
require_torch_multi_gpu,
slow,
)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
if model.config.model_type == "gpt2":
return model.transformer.h[0].mlp.c_fc
return model.transformer.h[0].mlp.dense_ah_to_h
if is_torch_available():
import torch
import torch.nn as nn
class A__ ( nn.Module ):
"""simple docstring"""
def __init__( self : List[str] , lowerCamelCase__ : nn.Module , lowerCamelCase__ : int ):
super().__init__()
a__ : int = module
a__ : Any = nn.Sequential(
nn.Linear(module.in_features , lowerCamelCase__ , bias=lowerCamelCase__ ) , nn.Linear(lowerCamelCase__ , module.out_features , bias=lowerCamelCase__ ) , )
a__ : Tuple = (2.0 / (5 * min(module.in_features , module.out_features ))) ** 0.5
nn.init.normal_(self.adapter[0].weight , std=lowerCamelCase__ )
nn.init.zeros_(self.adapter[1].weight )
self.adapter.to(module.weight.device )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Optional[int] , *lowerCamelCase__ : int , **lowerCamelCase__ : Dict ):
return self.module(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ ) + self.adapter(lowerCamelCase__ )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
_lowercase = 'bigscience/bloom-1b7'
# Constant values
_lowercase = 2.1_09_65_95_52_69_25_74
_lowercase = 'Hello my name is'
_lowercase = set()
EXPECTED_OUTPUTS.add('Hello my name is John and I am a professional photographer. I' )
EXPECTED_OUTPUTS.add('Hello my name is John.\nI am a friend of your father.\n' )
EXPECTED_OUTPUTS.add('Hello my name is John Doe, I am a student at the University' )
_lowercase = 1_0
def _UpperCamelCase( self : Dict ):
# Models and tokenizer
a__ : List[str] = AutoTokenizer.from_pretrained(self.model_name )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Union[str, Any] ):
super().setUp()
# Models and tokenizer
a__ : List[Any] = AutoModelForCausalLM.from_pretrained(
self.model_name , torch_dtype=torch.floataa , device_map="auto" )
a__ : Union[str, Any] = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.model_fpaa
del self.model_abit
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : List[Any] ):
a__ : str = self.model_abit.config
self.assertTrue(hasattr(lowerCamelCase__ , "quantization_config" ) )
a__ : Optional[Any] = config.to_dict()
a__ : int = config.to_diff_dict()
a__ : List[str] = config.to_json_string()
def _UpperCamelCase( self : int ):
from bitsandbytes.nn import Paramsabit
a__ : List[Any] = self.model_fpaa.get_memory_footprint()
a__ : str = self.model_abit.get_memory_footprint()
self.assertAlmostEqual(mem_fpaa / mem_abit , self.EXPECTED_RELATIVE_DIFFERENCE )
a__ : Optional[Any] = get_some_linear_layer(self.model_abit )
self.assertTrue(linear.weight.__class__ == Paramsabit )
def _UpperCamelCase( self : Tuple ):
from transformers import TaPreTrainedModel
self.model_fpaa.get_memory_footprint()
self.model_abit.get_memory_footprint()
for name, module in self.model_abit.named_modules():
if isinstance(lowerCamelCase__ , torch.nn.Linear ):
if name not in ["lm_head"] + TaPreTrainedModel._keep_in_fpaa_modules:
# 4-bit parameters are packed in uint8 variables
self.assertTrue(module.weight.dtype == torch.uinta )
def _UpperCamelCase( self : str ):
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Tuple = self.model_abit.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[Any] = BitsAndBytesConfig()
a__ : Optional[int] = True
a__ : int = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , device_map="auto" )
a__ : str = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : int = model_abit_from_config.generate(
input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : Dict ):
with self.assertRaises(lowerCamelCase__ ), tempfile.TemporaryDirectory() as tmpdirname:
self.model_abit.save_pretrained(lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : int = BitsAndBytesConfig()
with self.assertRaises(lowerCamelCase__ ):
a__ : Dict = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , load_in_abit=lowerCamelCase__ , device_map="auto" , bnb_abit_quant_type="nf4" , )
def _UpperCamelCase( self : int ):
with self.assertRaises(lowerCamelCase__ ):
# Tries with `str`
self.model_abit.to("cpu" )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `dtype``
self.model_abit.to(torch.floataa )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.to(torch.device("cuda:0" ) )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.float()
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.half()
# Test if we did not break anything
a__ : int = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Any = self.model_fpaa.to(torch.floataa )
a__ : List[Any] = self.model_fpaa.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.to("cpu" )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.half()
# Check this does not throw an error
a__ : Dict = self.model_fpaa.float()
def _UpperCamelCase( self : Dict ):
a__ : List[str] = AutoModelForSeqaSeqLM.from_pretrained("t5-small" , load_in_abit=lowerCamelCase__ , device_map="auto" )
self.assertTrue(model.decoder.block[0].layer[2].DenseReluDense.wo.weight.dtype == torch.floataa )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
@classmethod
def _UpperCamelCase( cls : str ):
a__ : Dict = "t5-small"
a__ : List[Any] = "google/flan-t5-small" # flan-t5 uses dense-act instead of dense-relu-dense
a__ : int = AutoTokenizer.from_pretrained(cls.model_name )
a__ : str = "Translate in German: Hello, my dog is cute"
def _UpperCamelCase( self : Optional[int] ):
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Optional[int] ):
from transformers import TaForConditionalGeneration
a__ : List[Any] = TaForConditionalGeneration._keep_in_fpaa_modules
a__ : Optional[Any] = None
# test with `t5-small`
a__ : Any = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : Dict = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Any = model.generate(**lowerCamelCase__ )
a__ : Union[str, Any] = modules
def _UpperCamelCase( self : List[Any] ):
import bitsandbytes as bnb
from transformers import TaForConditionalGeneration
# test with `t5-small`
a__ : List[str] = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# there was a bug with decoders - this test checks that it is fixed
self.assertTrue(isinstance(model.decoder.block[0].layer[0].SelfAttention.q , bnb.nn.Linearabit ) )
a__ : Union[str, Any] = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : int = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Any = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Optional[int] = model.generate(**lowerCamelCase__ )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : List[str] ):
super().setUp()
# model_name
a__ : Union[str, Any] = "bigscience/bloom-560m"
a__ : Union[str, Any] = "t5-small"
# Different types of model
a__ : int = AutoModel.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Sequence classification model
a__ : Dict = AutoModelForSequenceClassification.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# CausalLM model
a__ : str = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Seq2seq model
a__ : Dict = AutoModelForSeqaSeqLM.from_pretrained(
self.seq_to_seq_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.base_model
del self.sequence_model
del self.model_abit
del self.seq_to_seq_model
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Union[str, Any] ):
from bitsandbytes.nn import Paramsabit
self.assertTrue(self.base_model.h[-1].mlp.dense_ah_to_h.weight.__class__ == Paramsabit )
# Other heads should be nn.Parameter
self.assertTrue(self.model_abit.lm_head.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.sequence_model.score.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.seq_to_seq_model.lm_head.weight.__class__ == torch.nn.Parameter )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
super().setUp()
def _UpperCamelCase( self : int ):
del self.pipe
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Tuple ):
a__ : int = pipeline(
"text-generation" , model=self.model_name , model_kwargs={"device_map": "auto", "load_in_4bit": True, "torch_dtype": torch.floataa} , max_new_tokens=self.MAX_NEW_TOKENS , )
# Real second forward pass
a__ : Tuple = self.pipe(self.input_text )
self.assertIn(pipeline_output[0]["generated_text"] , self.EXPECTED_OUTPUTS )
@require_torch_multi_gpu
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Tuple ):
super().setUp()
def _UpperCamelCase( self : List[Any] ):
a__ : str = AutoModelForCausalLM.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="balanced" )
# Check correct device map
self.assertEqual(set(model_parallel.hf_device_map.values() ) , {0, 1} )
# Check that inference pass works on the model
a__ : List[Any] = self.tokenizer(self.input_text , return_tensors="pt" )
# Second real batch
a__ : List[Any] = model_parallel.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_parallel[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
a__ : Any = "facebook/opt-350m"
super().setUp()
def _UpperCamelCase( self : int ):
if version.parse(importlib.metadata.version("bitsandbytes" ) ) < version.parse("0.37.0" ):
return
# Step 1: freeze all parameters
a__ : Tuple = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ )
self.assertEqual(set(model.hf_device_map.values() ) , {torch.cuda.current_device()} )
for param in model.parameters():
a__ : Any = False # freeze the model - train adapters later
if param.ndim == 1:
# cast the small parameters (e.g. layernorm) to fp32 for stability
a__ : Tuple = param.data.to(torch.floataa )
# Step 2: add adapters
for _, module in model.named_modules():
if "OPTAttention" in repr(type(lowerCamelCase__ ) ):
a__ : Dict = LoRALayer(module.q_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.k_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.v_proj , rank=16 )
# Step 3: dummy batch
a__ : Dict = self.tokenizer("Test batch " , return_tensors="pt" ).to(0 )
# Step 4: Check if the gradient is not None
with torch.cuda.amp.autocast():
a__ : Optional[Any] = model.forward(**lowerCamelCase__ )
out.logits.norm().backward()
for module in model.modules():
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertTrue(module.adapter[1].weight.grad is not None )
self.assertTrue(module.adapter[1].weight.grad.norm().item() > 0 )
elif isinstance(lowerCamelCase__ , nn.Embedding ):
self.assertTrue(module.weight.grad is None )
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gpt2-xl'
_lowercase = 3.31_91_85_48_54_15_21_87
| 37 | 1 |
UpperCamelCase : Optional[int] = """Input must be a string of 8 numbers plus letter"""
UpperCamelCase : Tuple = """TRWAGMYFPDXBNJZSQVHLCKE"""
def UpperCamelCase_ ( __a ) -> bool:
if not isinstance(__a , __a ):
a__ : Dict = f'''Expected string as input, found {type(__a ).__name__}'''
raise TypeError(__a )
a__ : str = spanish_id.replace("-" , "" ).upper()
if len(__a ) != 9:
raise ValueError(__a )
try:
a__ : List[str] = int(spanish_id_clean[0:8] )
a__ : Any = spanish_id_clean[8]
except ValueError as ex:
raise ValueError(__a ) from ex
if letter.isdigit():
raise ValueError(__a )
return letter == LOOKUP_LETTERS[number % 23]
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 |
import inspect
import unittest
from datasets import load_dataset
from packaging import version
from transformers import BeitConfig
from transformers.models.auto import get_values
from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_vision_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, _config_zero_init, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import (
MODEL_MAPPING,
BeitForImageClassification,
BeitForMaskedImageModeling,
BeitForSemanticSegmentation,
BeitModel,
)
from transformers.models.beit.modeling_beit import BEIT_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
import PIL
from PIL import Image
from transformers import BeitImageProcessor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Optional[int]=100 , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[int]=30 , lowerCamelCase__ : Union[str, Any]=2 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : List[Any]=True , lowerCamelCase__ : Tuple=True , lowerCamelCase__ : int=32 , lowerCamelCase__ : Union[str, Any]=4 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Union[str, Any]=37 , lowerCamelCase__ : List[Any]="gelu" , lowerCamelCase__ : List[Any]=0.1 , lowerCamelCase__ : int=0.1 , lowerCamelCase__ : Union[str, Any]=10 , lowerCamelCase__ : str=0.02 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]=[0, 1, 2, 3] , ):
a__ : Dict = parent
a__ : Dict = 100
a__ : Optional[int] = batch_size
a__ : Union[str, Any] = image_size
a__ : Any = patch_size
a__ : Optional[Any] = num_channels
a__ : int = is_training
a__ : List[str] = use_labels
a__ : Optional[Any] = hidden_size
a__ : List[Any] = num_hidden_layers
a__ : str = num_attention_heads
a__ : str = intermediate_size
a__ : int = hidden_act
a__ : List[Any] = hidden_dropout_prob
a__ : Dict = attention_probs_dropout_prob
a__ : Union[str, Any] = type_sequence_label_size
a__ : Optional[Any] = initializer_range
a__ : List[str] = scope
a__ : int = out_indices
a__ : List[str] = num_labels
# in BeiT, the seq length equals the number of patches + 1 (we add 1 for the [CLS] token)
a__ : Optional[int] = (image_size // patch_size) ** 2
a__ : Union[str, Any] = num_patches + 1
def _UpperCamelCase( self : int ):
a__ : Optional[int] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
a__ : Optional[Any] = None
a__ : Tuple = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : Dict = ids_tensor([self.batch_size, self.image_size, self.image_size] , self.num_labels )
a__ : Optional[int] = self.get_config()
return config, pixel_values, labels, pixel_labels
def _UpperCamelCase( self : Tuple ):
return BeitConfig(
vocab_size=self.vocab_size , image_size=self.image_size , patch_size=self.patch_size , num_channels=self.num_channels , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , out_indices=self.out_indices , )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : int , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any ):
a__ : str = BeitModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[str] , lowerCamelCase__ : Any , lowerCamelCase__ : List[str] , lowerCamelCase__ : Tuple ):
a__ : int = BeitForMaskedImageModeling(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[Any] = model(lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length - 1, self.vocab_size) )
def _UpperCamelCase( self : str , lowerCamelCase__ : Any , lowerCamelCase__ : str , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Any ):
a__ : List[str] = self.type_sequence_label_size
a__ : Optional[Any] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
# test greyscale images
a__ : Optional[Any] = 1
a__ : List[str] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = floats_tensor([self.batch_size, 1, self.image_size, self.image_size] )
a__ : Union[str, Any] = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def _UpperCamelCase( self : Any , lowerCamelCase__ : str , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
a__ : int = self.num_labels
a__ : List[str] = BeitForSemanticSegmentation(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Tuple = model(lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
def _UpperCamelCase( self : Optional[int] ):
a__ : Any = self.prepare_config_and_inputs()
a__, a__, a__, a__ : Union[str, Any] = config_and_inputs
a__ : Dict = {"pixel_values": pixel_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(BeitModel, BeitForImageClassification, BeitForMaskedImageModeling, BeitForSemanticSegmentation)
if is_torch_available()
else ()
)
_lowercase = (
{
'feature-extraction': BeitModel,
'image-classification': BeitForImageClassification,
'image-segmentation': BeitForSemanticSegmentation,
}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Any ):
a__ : int = BeitModelTester(self )
a__ : Optional[Any] = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[Any] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="BEiT does not use inputs_embeds" )
def _UpperCamelCase( self : str ):
pass
@require_torch_multi_gpu
@unittest.skip(reason="BEiT has some layers using `add_module` which doesn't work well with `nn.DataParallel`" )
def _UpperCamelCase( self : Dict ):
pass
def _UpperCamelCase( self : Optional[Any] ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : List[str] = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Optional[int] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : str ):
a__, a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : int = model_class(lowerCamelCase__ )
a__ : str = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Any = ["pixel_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
def _UpperCamelCase( self : int ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_lm(*lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] ):
a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_semantic_segmentation(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
if not self.model_tester.is_training:
return
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : str = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]:
continue
a__ : List[str] = model_class(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.train()
a__ : Any = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : Tuple = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : Tuple ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
if not self.model_tester.is_training:
return
a__ : List[Any] = False
a__ : List[str] = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if (
model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]
or not model_class.supports_gradient_checkpointing
):
continue
a__ : Optional[Any] = model_class(lowerCamelCase__ )
model.gradient_checkpointing_enable()
model.to(lowerCamelCase__ )
model.train()
a__ : Union[str, Any] = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : int = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : List[str] ):
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : Dict = _config_zero_init(lowerCamelCase__ )
for model_class in self.all_model_classes:
a__ : str = model_class(config=lowerCamelCase__ )
for name, param in model.named_parameters():
# we skip lambda parameters as these require special initial values
# determined by config.layer_scale_init_value
if "lambda" in name:
continue
if param.requires_grad:
self.assertIn(
((param.data.mean() * 1E9).round() / 1E9).item() , [0.0, 1.0] , msg=f'''Parameter {name} of model {model_class} seems not properly initialized''' , )
@slow
def _UpperCamelCase( self : Optional[int] ):
for model_name in BEIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Tuple = BeitModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : List[Any] = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
return image
@require_torch
@require_vision
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : Optional[int] ):
return BeitImageProcessor.from_pretrained("microsoft/beit-base-patch16-224" ) if is_vision_available() else None
@slow
def _UpperCamelCase( self : str ):
a__ : int = BeitForMaskedImageModeling.from_pretrained("microsoft/beit-base-patch16-224-pt22k" ).to(lowerCamelCase__ )
a__ : Optional[Any] = self.default_image_processor
a__ : Dict = prepare_img()
a__ : Optional[int] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).pixel_values.to(lowerCamelCase__ )
# prepare bool_masked_pos
a__ : Optional[Any] = torch.ones((1, 196) , dtype=torch.bool ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(pixel_values=lowerCamelCase__ , bool_masked_pos=lowerCamelCase__ )
a__ : Tuple = outputs.logits
# verify the logits
a__ : List[str] = torch.Size((1, 196, 8_192) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[int] = torch.tensor(
[[-3.2437, 0.5072, -13.9174], [-3.2456, 0.4948, -13.9401], [-3.2033, 0.5121, -13.8550]] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[bool_masked_pos][:3, :3] , lowerCamelCase__ , atol=1E-2 ) )
@slow
def _UpperCamelCase( self : Dict ):
a__ : str = BeitForImageClassification.from_pretrained("microsoft/beit-base-patch16-224" ).to(lowerCamelCase__ )
a__ : int = self.default_image_processor
a__ : List[Any] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Union[str, Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 1_000) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = torch.tensor([-1.2385, -1.0987, -1.0108] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Tuple = 281
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : Any ):
a__ : Dict = BeitForImageClassification.from_pretrained("microsoft/beit-large-patch16-224-pt22k-ft22k" ).to(
lowerCamelCase__ )
a__ : str = self.default_image_processor
a__ : List[str] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Dict = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Optional[int] = torch.Size((1, 21_841) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[Any] = torch.tensor([1.6881, -0.2787, 0.5901] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Optional[Any] = 2_396
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
a__ : Optional[Any] = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : Tuple = model.to(lowerCamelCase__ )
a__ : List[Any] = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Tuple = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : Union[str, Any] = Image.open(ds[0]["file"] )
a__ : List[Any] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Optional[Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Tuple = torch.Size((1, 150, 160, 160) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = version.parse(PIL.__version__ ) < version.parse("9.0.0" )
if is_pillow_less_than_a:
a__ : Dict = torch.tensor(
[
[[-4.9225, -2.3954, -3.0522], [-2.8822, -1.0046, -1.7561], [-2.9549, -1.3228, -2.1347]],
[[-5.8168, -3.4129, -4.0778], [-3.8651, -2.2214, -3.0277], [-3.8356, -2.4643, -3.3535]],
[[-0.0078, 3.9952, 4.0754], [2.9856, 4.6944, 5.0035], [3.2413, 4.7813, 4.9969]],
] , device=lowerCamelCase__ , )
else:
a__ : Dict = torch.tensor(
[
[[-4.8960, -2.3688, -3.0355], [-2.8478, -0.9836, -1.7418], [-2.9449, -1.3332, -2.1456]],
[[-5.8081, -3.4124, -4.1006], [-3.8561, -2.2081, -3.0323], [-3.8365, -2.4601, -3.3669]],
[[-0.0309, 3.9868, 4.0540], [2.9640, 4.6877, 4.9976], [3.2081, 4.7690, 4.9942]],
] , device=lowerCamelCase__ , )
self.assertTrue(torch.allclose(logits[0, :3, :3, :3] , lowerCamelCase__ , atol=1E-4 ) )
@slow
def _UpperCamelCase( self : Tuple ):
a__ : str = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : List[Any] = model.to(lowerCamelCase__ )
a__ : int = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Optional[int] = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : str = Image.open(ds[0]["file"] )
a__ : str = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : List[Any] = model(**lowerCamelCase__ )
a__ : Any = outputs.logits.detach().cpu()
a__ : List[Any] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ , target_sizes=[(500, 300)] )
a__ : Optional[int] = torch.Size((500, 300) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
a__ : List[str] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ )
a__ : Any = torch.Size((160, 160) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
| 37 | 1 |
import argparse
import json
from pathlib import Path
import requests
import torch
from huggingface_hub import hf_hub_download
from PIL import Image
from timm import create_model
from timm.data import resolve_data_config
from timm.data.transforms_factory import create_transform
from transformers import BitConfig, BitForImageClassification, BitImageProcessor
from transformers.image_utils import PILImageResampling
from transformers.utils import logging
logging.set_verbosity_info()
UpperCamelCase : List[str] = logging.get_logger(__name__)
def UpperCamelCase_ ( __a ) -> str:
a__ : Optional[Any] = "huggingface/label-files"
a__ : Tuple = "imagenet-1k-id2label.json"
a__ : Dict = json.load(open(hf_hub_download(__a , __a , repo_type="dataset" ) , "r" ) )
a__ : List[str] = {int(__a ): v for k, v in idalabel.items()}
a__ : List[Any] = {v: k for k, v in idalabel.items()}
a__ : Dict = "std_conv" if "bit" in model_name else False
# note that when using BiT as backbone for ViT-hybrid checkpoints,
# one needs to additionally set config.layer_type = "bottleneck", config.stem_type = "same",
# config.conv_layer = "std_conv_same"
a__ : Tuple = BitConfig(
conv_layer=__a , num_labels=1_000 , idalabel=__a , labelaid=__a , )
return config
def UpperCamelCase_ ( __a ) -> Any:
if "stem.conv" in name:
a__ : Dict = name.replace("stem.conv" , "bit.embedder.convolution" )
if "blocks" in name:
a__ : str = name.replace("blocks" , "layers" )
if "head.fc" in name:
a__ : Union[str, Any] = name.replace("head.fc" , "classifier.1" )
if name.startswith("norm" ):
a__ : List[Any] = "bit." + name
if "bit" not in name and "classifier" not in name:
a__ : List[Any] = "bit.encoder." + name
return name
def UpperCamelCase_ ( ) -> Optional[Any]:
a__ : Dict = "http://images.cocodataset.org/val2017/000000039769.jpg"
a__ : Optional[Any] = Image.open(requests.get(__a , stream=__a ).raw )
return im
@torch.no_grad()
def UpperCamelCase_ ( __a , __a , __a=False ) -> Any:
a__ : Optional[int] = get_config(__a )
# load original model from timm
a__ : List[str] = create_model(__a , pretrained=__a )
timm_model.eval()
# load state_dict of original model
a__ : Optional[int] = timm_model.state_dict()
for key in state_dict.copy().keys():
a__ : Tuple = state_dict.pop(__a )
a__ : str = val.squeeze() if "head" in key else val
# load HuggingFace model
a__ : str = BitForImageClassification(__a )
model.eval()
model.load_state_dict(__a )
# create image processor
a__ : List[Any] = create_transform(**resolve_data_config({} , model=__a ) )
a__ : Any = transform.transforms
a__ : Optional[int] = {
"bilinear": PILImageResampling.BILINEAR,
"bicubic": PILImageResampling.BICUBIC,
"nearest": PILImageResampling.NEAREST,
}
a__ : str = BitImageProcessor(
do_resize=__a , size={"shortest_edge": timm_transforms[0].size} , resample=pillow_resamplings[timm_transforms[0].interpolation.value] , do_center_crop=__a , crop_size={"height": timm_transforms[1].size[0], "width": timm_transforms[1].size[1]} , do_normalize=__a , image_mean=timm_transforms[-1].mean.tolist() , image_std=timm_transforms[-1].std.tolist() , )
a__ : Tuple = prepare_img()
a__ : Dict = transform(__a ).unsqueeze(0 )
a__ : List[Any] = processor(__a , return_tensors="pt" ).pixel_values
# verify pixel values
assert torch.allclose(__a , __a )
# verify logits
with torch.no_grad():
a__ : int = model(__a )
a__ : str = outputs.logits
print("Logits:" , logits[0, :3] )
print("Predicted class:" , model.config.idalabel[logits.argmax(-1 ).item()] )
a__ : Tuple = timm_model(__a )
assert timm_logits.shape == outputs.logits.shape
assert torch.allclose(__a , outputs.logits , atol=1e-3 )
print("Looks ok!" )
if pytorch_dump_folder_path is not None:
Path(__a ).mkdir(exist_ok=__a )
print(f'''Saving model {model_name} and processor to {pytorch_dump_folder_path}''' )
model.save_pretrained(__a )
processor.save_pretrained(__a )
if push_to_hub:
print(f'''Pushing model {model_name} and processor to the hub''' )
model.push_to_hub(f'''ybelkada/{model_name}''' )
processor.push_to_hub(f'''ybelkada/{model_name}''' )
if __name__ == "__main__":
UpperCamelCase : Dict = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"""--model_name""",
default="""resnetv2_50x1_bitm""",
type=str,
help="""Name of the BiT timm model you'd like to convert.""",
)
parser.add_argument(
"""--pytorch_dump_folder_path""", default=None, type=str, help="""Path to the output PyTorch model directory."""
)
parser.add_argument(
"""--push_to_hub""",
action="""store_true""",
help="""Whether to push the model to the hub.""",
)
UpperCamelCase : List[Any] = parser.parse_args()
convert_bit_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 37 |
import re
import jax.numpy as jnp
from flax.traverse_util import flatten_dict, unflatten_dict
from jax.random import PRNGKey
from ..utils import logging
UpperCamelCase : Dict = logging.get_logger(__name__)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
a__ : Tuple = R"\w+[.]\d+"
a__ : List[Any] = re.findall(__a , __a )
for pat in pats:
a__ : Union[str, Any] = key.replace(__a , "_".join(pat.split("." ) ) )
return key
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : List[str] = pt_tuple_key[:-1] + ("scale",)
if (
any("norm" in str_ for str_ in pt_tuple_key )
and (pt_tuple_key[-1] == "bias")
and (pt_tuple_key[:-1] + ("bias",) not in random_flax_state_dict)
and (pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict)
):
a__ : Any = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
elif pt_tuple_key[-1] in ["weight", "gamma"] and pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict:
a__ : Optional[Any] = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
# embedding
if pt_tuple_key[-1] == "weight" and pt_tuple_key[:-1] + ("embedding",) in random_flax_state_dict:
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("embedding",)
return renamed_pt_tuple_key, pt_tensor
# conv layer
a__ : List[str] = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight" and pt_tensor.ndim == 4:
a__ : str = pt_tensor.transpose(2 , 3 , 1 , 0 )
return renamed_pt_tuple_key, pt_tensor
# linear layer
a__ : Tuple = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight":
a__ : Tuple = pt_tensor.T
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm weight
a__ : Optional[Any] = pt_tuple_key[:-1] + ("weight",)
if pt_tuple_key[-1] == "gamma":
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm bias
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("bias",)
if pt_tuple_key[-1] == "beta":
return renamed_pt_tuple_key, pt_tensor
return pt_tuple_key, pt_tensor
def UpperCamelCase_ ( __a , __a , __a=42 ) -> str:
# Step 1: Convert pytorch tensor to numpy
a__ : Optional[int] = {k: v.numpy() for k, v in pt_state_dict.items()}
# Step 2: Since the model is stateless, get random Flax params
a__ : Tuple = flax_model.init_weights(PRNGKey(__a ) )
a__ : Optional[Any] = flatten_dict(__a )
a__ : Union[str, Any] = {}
# Need to change some parameters name to match Flax names
for pt_key, pt_tensor in pt_state_dict.items():
a__ : Optional[int] = rename_key(__a )
a__ : Optional[int] = tuple(renamed_pt_key.split("." ) )
# Correctly rename weight parameters
a__, a__ : Union[str, Any] = rename_key_and_reshape_tensor(__a , __a , __a )
if flax_key in random_flax_state_dict:
if flax_tensor.shape != random_flax_state_dict[flax_key].shape:
raise ValueError(
f'''PyTorch checkpoint seems to be incorrect. Weight {pt_key} was expected to be of shape '''
f'''{random_flax_state_dict[flax_key].shape}, but is {flax_tensor.shape}.''' )
# also add unexpected weight so that warning is thrown
a__ : str = jnp.asarray(__a )
return unflatten_dict(__a )
| 37 | 1 |
import json
import sys
import tempfile
import unittest
from pathlib import Path
import transformers
from transformers import (
CONFIG_MAPPING,
FEATURE_EXTRACTOR_MAPPING,
AutoConfig,
AutoFeatureExtractor,
WavaVecaConfig,
WavaVecaFeatureExtractor,
)
from transformers.testing_utils import DUMMY_UNKNOWN_IDENTIFIER, get_tests_dir
sys.path.append(str(Path(__file__).parent.parent.parent.parent / """utils"""))
from test_module.custom_configuration import CustomConfig # noqa E402
from test_module.custom_feature_extraction import CustomFeatureExtractor # noqa E402
UpperCamelCase : Optional[int] = get_tests_dir("""fixtures""")
UpperCamelCase : List[Any] = get_tests_dir("""fixtures/dummy_feature_extractor_config.json""")
UpperCamelCase : List[str] = get_tests_dir("""fixtures/dummy-config.json""")
class A__ ( unittest.TestCase ):
"""simple docstring"""
def _UpperCamelCase( self : str ):
a__ : int = 0
def _UpperCamelCase( self : Dict ):
a__ : Any = AutoFeatureExtractor.from_pretrained("facebook/wav2vec2-base-960h" )
self.assertIsInstance(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : str ):
a__ : Union[str, Any] = AutoFeatureExtractor.from_pretrained(lowerCamelCase__ )
self.assertIsInstance(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
with tempfile.TemporaryDirectory() as tmpdirname:
a__ : Optional[int] = WavaVecaConfig()
# remove feature_extractor_type to make sure config.json alone is enough to load feature processor locally
a__ : Union[str, Any] = AutoFeatureExtractor.from_pretrained(lowerCamelCase__ ).to_dict()
config_dict.pop("feature_extractor_type" )
a__ : Tuple = WavaVecaFeatureExtractor(**lowerCamelCase__ )
# save in new folder
model_config.save_pretrained(lowerCamelCase__ )
config.save_pretrained(lowerCamelCase__ )
a__ : Dict = AutoFeatureExtractor.from_pretrained(lowerCamelCase__ )
# make sure private variable is not incorrectly saved
a__ : Dict = json.loads(config.to_json_string() )
self.assertTrue("_processor_class" not in dict_as_saved )
self.assertIsInstance(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] ):
a__ : Union[str, Any] = AutoFeatureExtractor.from_pretrained(lowerCamelCase__ )
self.assertIsInstance(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Any ):
with self.assertRaisesRegex(
lowerCamelCase__ , "bert-base is not a local folder and is not a valid model identifier" ):
a__ : Tuple = AutoFeatureExtractor.from_pretrained("bert-base" )
def _UpperCamelCase( self : List[Any] ):
with self.assertRaisesRegex(
lowerCamelCase__ , r"aaaaaa is not a valid git identifier \(branch name, tag name or commit id\)" ):
a__ : Union[str, Any] = AutoFeatureExtractor.from_pretrained(lowerCamelCase__ , revision="aaaaaa" )
def _UpperCamelCase( self : Union[str, Any] ):
with self.assertRaisesRegex(
lowerCamelCase__ , "hf-internal-testing/config-no-model does not appear to have a file named preprocessor_config.json." , ):
a__ : List[Any] = AutoFeatureExtractor.from_pretrained("hf-internal-testing/config-no-model" )
def _UpperCamelCase( self : Union[str, Any] ):
# If remote code is not set, we will time out when asking whether to load the model.
with self.assertRaises(lowerCamelCase__ ):
a__ : str = AutoFeatureExtractor.from_pretrained(
"hf-internal-testing/test_dynamic_feature_extractor" )
# If remote code is disabled, we can't load this config.
with self.assertRaises(lowerCamelCase__ ):
a__ : Tuple = AutoFeatureExtractor.from_pretrained(
"hf-internal-testing/test_dynamic_feature_extractor" , trust_remote_code=lowerCamelCase__ )
a__ : List[str] = AutoFeatureExtractor.from_pretrained(
"hf-internal-testing/test_dynamic_feature_extractor" , trust_remote_code=lowerCamelCase__ )
self.assertEqual(feature_extractor.__class__.__name__ , "NewFeatureExtractor" )
# Test feature extractor can be reloaded.
with tempfile.TemporaryDirectory() as tmp_dir:
feature_extractor.save_pretrained(lowerCamelCase__ )
a__ : Tuple = AutoFeatureExtractor.from_pretrained(lowerCamelCase__ , trust_remote_code=lowerCamelCase__ )
self.assertEqual(reloaded_feature_extractor.__class__.__name__ , "NewFeatureExtractor" )
def _UpperCamelCase( self : Optional[Any] ):
try:
AutoConfig.register("custom" , lowerCamelCase__ )
AutoFeatureExtractor.register(lowerCamelCase__ , lowerCamelCase__ )
# Trying to register something existing in the Transformers library will raise an error
with self.assertRaises(lowerCamelCase__ ):
AutoFeatureExtractor.register(lowerCamelCase__ , lowerCamelCase__ )
# Now that the config is registered, it can be used as any other config with the auto-API
a__ : Optional[Any] = CustomFeatureExtractor.from_pretrained(lowerCamelCase__ )
with tempfile.TemporaryDirectory() as tmp_dir:
feature_extractor.save_pretrained(lowerCamelCase__ )
a__ : Dict = AutoFeatureExtractor.from_pretrained(lowerCamelCase__ )
self.assertIsInstance(lowerCamelCase__ , lowerCamelCase__ )
finally:
if "custom" in CONFIG_MAPPING._extra_content:
del CONFIG_MAPPING._extra_content["custom"]
if CustomConfig in FEATURE_EXTRACTOR_MAPPING._extra_content:
del FEATURE_EXTRACTOR_MAPPING._extra_content[CustomConfig]
def _UpperCamelCase( self : int ):
class A__ ( A__ ):
"""simple docstring"""
_lowercase = True
try:
AutoConfig.register("custom" , lowerCamelCase__ )
AutoFeatureExtractor.register(lowerCamelCase__ , lowerCamelCase__ )
# If remote code is not set, the default is to use local
a__ : Dict = AutoFeatureExtractor.from_pretrained(
"hf-internal-testing/test_dynamic_feature_extractor" )
self.assertEqual(feature_extractor.__class__.__name__ , "NewFeatureExtractor" )
self.assertTrue(feature_extractor.is_local )
# If remote code is disabled, we load the local one.
a__ : List[str] = AutoFeatureExtractor.from_pretrained(
"hf-internal-testing/test_dynamic_feature_extractor" , trust_remote_code=lowerCamelCase__ )
self.assertEqual(feature_extractor.__class__.__name__ , "NewFeatureExtractor" )
self.assertTrue(feature_extractor.is_local )
# If remote is enabled, we load from the Hub
a__ : int = AutoFeatureExtractor.from_pretrained(
"hf-internal-testing/test_dynamic_feature_extractor" , trust_remote_code=lowerCamelCase__ )
self.assertEqual(feature_extractor.__class__.__name__ , "NewFeatureExtractor" )
self.assertTrue(not hasattr(lowerCamelCase__ , "is_local" ) )
finally:
if "custom" in CONFIG_MAPPING._extra_content:
del CONFIG_MAPPING._extra_content["custom"]
if CustomConfig in FEATURE_EXTRACTOR_MAPPING._extra_content:
del FEATURE_EXTRACTOR_MAPPING._extra_content[CustomConfig]
| 37 |
from transformers import HfArgumentParser, TensorFlowBenchmark, TensorFlowBenchmarkArguments
def UpperCamelCase_ ( ) -> int:
a__ : Any = HfArgumentParser(__a )
a__ : Any = parser.parse_args_into_dataclasses()[0]
a__ : Optional[int] = TensorFlowBenchmark(args=__a )
try:
a__ : Optional[int] = parser.parse_args_into_dataclasses()[0]
except ValueError as e:
a__ : Tuple = "Arg --no_{0} is no longer used, please use --no-{0} instead."
a__ : List[Any] = " ".join(str(__a ).split(" " )[:-1] )
a__ : str = ""
a__ : List[Any] = eval(str(__a ).split(" " )[-1] )
a__ : List[str] = []
for arg in depreciated_args:
# arg[2:] removes '--'
if arg[2:] in TensorFlowBenchmark.deprecated_args:
# arg[5:] removes '--no_'
full_error_msg += arg_error_msg.format(arg[5:] )
else:
wrong_args.append(__a )
if len(__a ) > 0:
a__ : Tuple = full_error_msg + begin_error_msg + str(__a )
raise ValueError(__a )
benchmark.run()
if __name__ == "__main__":
main()
| 37 | 1 |
import unittest
from transformers import XLMConfig, is_torch_available
from transformers.testing_utils import require_torch, slow, torch_device
from ...generation.test_utils import GenerationTesterMixin
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from transformers import (
XLMForMultipleChoice,
XLMForQuestionAnswering,
XLMForQuestionAnsweringSimple,
XLMForSequenceClassification,
XLMForTokenClassification,
XLMModel,
XLMWithLMHeadModel,
)
from transformers.models.xlm.modeling_xlm import XLM_PRETRAINED_MODEL_ARCHIVE_LIST
class A__ :
"""simple docstring"""
def __init__( self : Any , lowerCamelCase__ : str , lowerCamelCase__ : str=13 , lowerCamelCase__ : int=7 , lowerCamelCase__ : Dict=True , lowerCamelCase__ : Optional[int]=True , lowerCamelCase__ : Union[str, Any]=True , lowerCamelCase__ : Any=True , lowerCamelCase__ : Tuple=True , lowerCamelCase__ : Dict=False , lowerCamelCase__ : Optional[int]=False , lowerCamelCase__ : Tuple=False , lowerCamelCase__ : int=2 , lowerCamelCase__ : Union[str, Any]=99 , lowerCamelCase__ : Optional[Any]=0 , lowerCamelCase__ : Union[str, Any]=32 , lowerCamelCase__ : Any=5 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Optional[Any]=0.1 , lowerCamelCase__ : Dict=0.1 , lowerCamelCase__ : Union[str, Any]=512 , lowerCamelCase__ : Union[str, Any]=2 , lowerCamelCase__ : int=0.02 , lowerCamelCase__ : List[Any]=2 , lowerCamelCase__ : Optional[Any]=4 , lowerCamelCase__ : List[Any]="last" , lowerCamelCase__ : Dict=True , lowerCamelCase__ : Union[str, Any]=None , lowerCamelCase__ : str=0 , ):
a__ : int = parent
a__ : Optional[Any] = batch_size
a__ : Optional[int] = seq_length
a__ : Union[str, Any] = is_training
a__ : List[str] = use_input_lengths
a__ : Optional[Any] = use_token_type_ids
a__ : Optional[int] = use_labels
a__ : List[Any] = gelu_activation
a__ : List[Any] = sinusoidal_embeddings
a__ : Optional[int] = causal
a__ : Dict = asm
a__ : str = n_langs
a__ : Optional[Any] = vocab_size
a__ : Union[str, Any] = n_special
a__ : List[Any] = hidden_size
a__ : Dict = num_hidden_layers
a__ : str = num_attention_heads
a__ : Union[str, Any] = hidden_dropout_prob
a__ : str = attention_probs_dropout_prob
a__ : int = max_position_embeddings
a__ : List[Any] = type_sequence_label_size
a__ : List[str] = initializer_range
a__ : Union[str, Any] = num_labels
a__ : str = num_choices
a__ : Union[str, Any] = summary_type
a__ : Tuple = use_proj
a__ : int = scope
a__ : Any = bos_token_id
def _UpperCamelCase( self : Any ):
a__ : Optional[Any] = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
a__ : Any = random_attention_mask([self.batch_size, self.seq_length] )
a__ : List[Any] = None
if self.use_input_lengths:
a__ : int = (
ids_tensor([self.batch_size] , vocab_size=2 ) + self.seq_length - 2
) # small variation of seq_length
a__ : List[Any] = None
if self.use_token_type_ids:
a__ : Optional[Any] = ids_tensor([self.batch_size, self.seq_length] , self.n_langs )
a__ : Optional[Any] = None
a__ : Tuple = None
a__ : str = None
if self.use_labels:
a__ : List[Any] = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : int = ids_tensor([self.batch_size, self.seq_length] , self.num_labels )
a__ : str = ids_tensor([self.batch_size] , 2 ).float()
a__ : str = ids_tensor([self.batch_size] , self.num_choices )
a__ : Tuple = self.get_config()
return (
config,
input_ids,
token_type_ids,
input_lengths,
sequence_labels,
token_labels,
is_impossible_labels,
choice_labels,
input_mask,
)
def _UpperCamelCase( self : Dict ):
return XLMConfig(
vocab_size=self.vocab_size , n_special=self.n_special , emb_dim=self.hidden_size , n_layers=self.num_hidden_layers , n_heads=self.num_attention_heads , dropout=self.hidden_dropout_prob , attention_dropout=self.attention_probs_dropout_prob , gelu_activation=self.gelu_activation , sinusoidal_embeddings=self.sinusoidal_embeddings , asm=self.asm , causal=self.causal , n_langs=self.n_langs , max_position_embeddings=self.max_position_embeddings , initializer_range=self.initializer_range , summary_type=self.summary_type , use_proj=self.use_proj , num_labels=self.num_labels , bos_token_id=self.bos_token_id , )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : int , lowerCamelCase__ : List[str] , lowerCamelCase__ : Any , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : str , ):
a__ : Any = XLMModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = model(lowerCamelCase__ , lengths=lowerCamelCase__ , langs=lowerCamelCase__ )
a__ : Optional[int] = model(lowerCamelCase__ , langs=lowerCamelCase__ )
a__ : Tuple = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : str , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Any , lowerCamelCase__ : Any , ):
a__ : Dict = XLMWithLMHeadModel(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Optional[Any] = model(lowerCamelCase__ , token_type_ids=lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.vocab_size) )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : List[Any] , lowerCamelCase__ : int , lowerCamelCase__ : Dict , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[str] , lowerCamelCase__ : Tuple , lowerCamelCase__ : List[Any] , lowerCamelCase__ : List[Any] , ):
a__ : List[Any] = XLMForQuestionAnsweringSimple(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Tuple = model(lowerCamelCase__ )
a__ : str = model(lowerCamelCase__ , start_positions=lowerCamelCase__ , end_positions=lowerCamelCase__ )
a__ : Any = outputs
self.parent.assertEqual(result.start_logits.shape , (self.batch_size, self.seq_length) )
self.parent.assertEqual(result.end_logits.shape , (self.batch_size, self.seq_length) )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Any , lowerCamelCase__ : Tuple , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Dict , lowerCamelCase__ : Tuple , lowerCamelCase__ : Tuple , lowerCamelCase__ : Dict , ):
a__ : List[Any] = XLMForQuestionAnswering(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Optional[int] = model(lowerCamelCase__ )
a__ : Any = model(
lowerCamelCase__ , start_positions=lowerCamelCase__ , end_positions=lowerCamelCase__ , cls_index=lowerCamelCase__ , is_impossible=lowerCamelCase__ , p_mask=lowerCamelCase__ , )
a__ : Union[str, Any] = model(
lowerCamelCase__ , start_positions=lowerCamelCase__ , end_positions=lowerCamelCase__ , cls_index=lowerCamelCase__ , is_impossible=lowerCamelCase__ , )
((a__), ) : List[Any] = result_with_labels.to_tuple()
a__ : List[Any] = model(lowerCamelCase__ , start_positions=lowerCamelCase__ , end_positions=lowerCamelCase__ )
((a__), ) : List[Any] = result_with_labels.to_tuple()
self.parent.assertEqual(result_with_labels.loss.shape , () )
self.parent.assertEqual(result.start_top_log_probs.shape , (self.batch_size, model.config.start_n_top) )
self.parent.assertEqual(result.start_top_index.shape , (self.batch_size, model.config.start_n_top) )
self.parent.assertEqual(
result.end_top_log_probs.shape , (self.batch_size, model.config.start_n_top * model.config.end_n_top) )
self.parent.assertEqual(
result.end_top_index.shape , (self.batch_size, model.config.start_n_top * model.config.end_n_top) )
self.parent.assertEqual(result.cls_logits.shape , (self.batch_size,) )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : Tuple , lowerCamelCase__ : Tuple , lowerCamelCase__ : Dict , lowerCamelCase__ : List[str] , lowerCamelCase__ : Optional[Any] , ):
a__ : Optional[int] = XLMForSequenceClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Optional[int] = model(lowerCamelCase__ )
a__ : List[Any] = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.loss.shape , () )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def _UpperCamelCase( self : str , lowerCamelCase__ : Any , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : Tuple , ):
a__ : int = self.num_labels
a__ : Dict = XLMForTokenClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Dict = model(lowerCamelCase__ , attention_mask=lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length, self.num_labels) )
def _UpperCamelCase( self : int , lowerCamelCase__ : Any , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : int , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : str , ):
a__ : str = self.num_choices
a__ : Union[str, Any] = XLMForMultipleChoice(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Any = input_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
a__ : Optional[Any] = token_type_ids.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
a__ : Union[str, Any] = input_mask.unsqueeze(1 ).expand(-1 , self.num_choices , -1 ).contiguous()
a__ : Dict = model(
lowerCamelCase__ , attention_mask=lowerCamelCase__ , token_type_ids=lowerCamelCase__ , labels=lowerCamelCase__ , )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.num_choices) )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : str = self.prepare_config_and_inputs()
(
(
a__
), (
a__
), (
a__
), (
a__
), (
a__
), (
a__
), (
a__
), (
a__
), (
a__
),
) : Tuple = config_and_inputs
a__ : Dict = {"input_ids": input_ids, "token_type_ids": token_type_ids, "lengths": input_lengths}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(
XLMModel,
XLMWithLMHeadModel,
XLMForQuestionAnswering,
XLMForSequenceClassification,
XLMForQuestionAnsweringSimple,
XLMForTokenClassification,
XLMForMultipleChoice,
)
if is_torch_available()
else ()
)
_lowercase = (
(XLMWithLMHeadModel,) if is_torch_available() else ()
) # TODO (PVP): Check other models whether language generation is also applicable
_lowercase = (
{
'feature-extraction': XLMModel,
'fill-mask': XLMWithLMHeadModel,
'question-answering': XLMForQuestionAnsweringSimple,
'text-classification': XLMForSequenceClassification,
'text-generation': XLMWithLMHeadModel,
'token-classification': XLMForTokenClassification,
'zero-shot': XLMForSequenceClassification,
}
if is_torch_available()
else {}
)
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Dict , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Union[str, Any] ):
if (
pipeline_test_casse_name == "QAPipelineTests"
and tokenizer_name is not None
and not tokenizer_name.endswith("Fast" )
):
# `QAPipelineTests` fails for a few models when the slower tokenizer are used.
# (The slower tokenizers were never used for pipeline tests before the pipeline testing rework)
# TODO: check (and possibly fix) the `QAPipelineTests` with slower tokenizer
return True
return False
def _UpperCamelCase( self : int , lowerCamelCase__ : Dict , lowerCamelCase__ : str , lowerCamelCase__ : Optional[Any]=False ):
a__ : List[Any] = super()._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
if return_labels:
if model_class.__name__ == "XLMForQuestionAnswering":
a__ : int = torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=lowerCamelCase__ )
a__ : Optional[Any] = torch.zeros(
self.model_tester.batch_size , dtype=torch.long , device=lowerCamelCase__ )
return inputs_dict
def _UpperCamelCase( self : Dict ):
a__ : List[str] = XLMModelTester(self )
a__ : str = ConfigTester(self , config_class=lowerCamelCase__ , emb_dim=37 )
def _UpperCamelCase( self : Any ):
self.config_tester.run_common_tests()
def _UpperCamelCase( self : Tuple ):
a__ : List[str] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_model(*lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_lm_head(*lowerCamelCase__ )
def _UpperCamelCase( self : Any ):
a__ : Tuple = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_simple_qa(*lowerCamelCase__ )
def _UpperCamelCase( self : List[str] ):
a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_qa(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] ):
a__ : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_sequence_classif(*lowerCamelCase__ )
def _UpperCamelCase( self : int ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_token_classif(*lowerCamelCase__ )
def _UpperCamelCase( self : List[str] ):
a__ : List[str] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xlm_for_multiple_choice(*lowerCamelCase__ )
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : Any , lowerCamelCase__ : str , lowerCamelCase__ : Tuple , lowerCamelCase__ : Any , lowerCamelCase__ : List[str] , lowerCamelCase__ : Union[str, Any]=False , lowerCamelCase__ : Dict=1 ):
self.assertIsInstance(lowerCamelCase__ , lowerCamelCase__ )
self.assertListEqual(
[isinstance(lowerCamelCase__ , lowerCamelCase__ ) for iter_attentions in attentions] , [True] * len(lowerCamelCase__ ) )
self.assertEqual(len(lowerCamelCase__ ) , (max_length - min_length) * num_beam_groups )
for idx, iter_attentions in enumerate(lowerCamelCase__ ):
# adds PAD dummy token
a__ : Tuple = min_length + idx + 1
a__ : str = min_length + idx + 1
a__ : List[Any] = (
batch_size * num_beam_groups,
config.num_attention_heads,
tgt_len,
src_len,
)
# check attn size
self.assertListEqual(
[layer_attention.shape for layer_attention in iter_attentions] , [expected_shape] * len(lowerCamelCase__ ) )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Dict , lowerCamelCase__ : str , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : List[str]=False , lowerCamelCase__ : Optional[int]=1 ):
self.assertIsInstance(lowerCamelCase__ , lowerCamelCase__ )
self.assertListEqual(
[isinstance(lowerCamelCase__ , lowerCamelCase__ ) for iter_hidden_states in hidden_states] , [True] * len(lowerCamelCase__ ) , )
self.assertEqual(len(lowerCamelCase__ ) , (max_length - min_length) * num_beam_groups )
for idx, iter_hidden_states in enumerate(lowerCamelCase__ ):
# adds PAD dummy token
a__ : Tuple = min_length + idx + 1
a__ : List[Any] = (batch_size * num_beam_groups, seq_len, config.hidden_size)
# check hidden size
self.assertListEqual(
[layer_hidden_states.shape for layer_hidden_states in iter_hidden_states] , [expected_shape] * len(lowerCamelCase__ ) , )
pass
@slow
def _UpperCamelCase( self : Optional[Any] ):
for model_name in XLM_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Union[str, Any] = XLMModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
@require_torch
class A__ ( unittest.TestCase ):
"""simple docstring"""
@slow
def _UpperCamelCase( self : List[str] ):
a__ : Optional[Any] = XLMWithLMHeadModel.from_pretrained("xlm-mlm-en-2048" )
model.to(lowerCamelCase__ )
a__ : Any = torch.tensor([[14, 447]] , dtype=torch.long , device=lowerCamelCase__ ) # the president
a__ : Any = [
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
14,
447,
] # the president the president the president the president the president the president the president the president the president the president
# TODO(PVP): this and other input_ids I tried for generation give pretty bad results. Not sure why. Model might just not be made for auto-regressive inference
a__ : Optional[Any] = model.generate(lowerCamelCase__ , do_sample=lowerCamelCase__ )
self.assertListEqual(output_ids[0].cpu().numpy().tolist() , lowerCamelCase__ )
| 37 |
import argparse
import ast
import logging
import os
import sys
import pandas as pd
import torch
from tqdm import tqdm
from transformers import BartForConditionalGeneration, RagRetriever, RagSequenceForGeneration, RagTokenForGeneration
from transformers import logging as transformers_logging
sys.path.append(os.path.join(os.getcwd())) # noqa: E402 # isort:skip
from utils_rag import exact_match_score, fa_score # noqa: E402 # isort:skip
UpperCamelCase : Optional[int] = logging.getLogger(__name__)
logging.basicConfig(level=logging.INFO)
transformers_logging.set_verbosity_info()
def UpperCamelCase_ ( __a ) -> Any:
if "token" in model_name_or_path:
return "rag_token"
if "sequence" in model_name_or_path:
return "rag_sequence"
if "bart" in model_name_or_path:
return "bart"
return None
def UpperCamelCase_ ( __a , __a , __a ) -> Any:
return max(metric_fn(__a , __a ) for gt in ground_truths )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = []
if args.gold_data_mode == "qa":
a__ : Any = pd.read_csv(__a , sep="\t" , header=__a )
for answer_list in data[1]:
a__ : Union[str, Any] = ast.literal_eval(__a )
answers.append(__a )
else:
a__ : List[str] = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : List[str] = [[reference] for reference in references]
a__ : List[str] = 0
for prediction, ground_truths in zip(__a , __a ):
total += 1
em += metric_max_over_ground_truths(__a , __a , __a )
fa += metric_max_over_ground_truths(__a , __a , __a )
a__ : Dict = 100.0 * em / total
a__ : Optional[Any] = 100.0 * fa / total
logger.info(f'''F1: {fa:.2f}''' )
logger.info(f'''EM: {em:.2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = args.k
a__ : str = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = 0
for hypo, reference in zip(__a , __a ):
a__ : Any = set(hypo.split("\t" )[:k] )
a__ : Union[str, Any] = set(reference.split("\t" ) )
total += 1
em += len(hypo_provenance & ref_provenance ) / k
a__ : Union[str, Any] = 100.0 * em / total
logger.info(f'''Precision@{k}: {em: .2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
def strip_title(__a ):
if title.startswith("\"" ):
a__ : Optional[Any] = title[1:]
if title.endswith("\"" ):
a__ : Union[str, Any] = title[:-1]
return title
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a , )["input_ids"].to(args.device )
a__ : Optional[int] = rag_model.rag.question_encoder(__a )
a__ : Union[str, Any] = question_enc_outputs[0]
a__ : Optional[int] = rag_model.retriever(
__a , question_enc_pool_output.cpu().detach().to(torch.floataa ).numpy() , prefix=rag_model.rag.generator.config.prefix , n_docs=rag_model.config.n_docs , return_tensors="pt" , )
a__ : List[Any] = rag_model.retriever.index.get_doc_dicts(result.doc_ids )
a__ : int = []
for docs in all_docs:
a__ : Optional[int] = [strip_title(__a ) for title in docs["title"]]
provenance_strings.append("\t".join(__a ) )
return provenance_strings
def UpperCamelCase_ ( __a , __a , __a ) -> Dict:
with torch.no_grad():
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a )
a__ : Any = inputs_dict.input_ids.to(args.device )
a__ : Dict = inputs_dict.attention_mask.to(args.device )
a__ : Optional[int] = rag_model.generate( # rag_model overwrites generate
__a , attention_mask=__a , num_beams=args.num_beams , min_length=args.min_length , max_length=args.max_length , early_stopping=__a , num_return_sequences=1 , bad_words_ids=[[0, 0]] , )
a__ : int = rag_model.retriever.generator_tokenizer.batch_decode(__a , skip_special_tokens=__a )
if args.print_predictions:
for q, a in zip(__a , __a ):
logger.info("Q: {} - A: {}".format(__a , __a ) )
return answers
def UpperCamelCase_ ( ) -> List[str]:
a__ : int = argparse.ArgumentParser()
parser.add_argument(
"--model_type" , choices=["rag_sequence", "rag_token", "bart"] , type=__a , help=(
"RAG model type: rag_sequence, rag_token or bart, if none specified, the type is inferred from the"
" model_name_or_path"
) , )
parser.add_argument(
"--index_name" , default=__a , choices=["exact", "compressed", "legacy"] , type=__a , help="RAG model retriever type" , )
parser.add_argument(
"--index_path" , default=__a , type=__a , help="Path to the retrieval index" , )
parser.add_argument("--n_docs" , default=5 , type=__a , help="Number of retrieved docs" )
parser.add_argument(
"--model_name_or_path" , default=__a , type=__a , required=__a , help="Path to pretrained checkpoints or model identifier from huggingface.co/models" , )
parser.add_argument(
"--eval_mode" , choices=["e2e", "retrieval"] , default="e2e" , type=__a , help=(
"Evaluation mode, e2e calculates exact match and F1 of the downstream task, retrieval calculates"
" precision@k."
) , )
parser.add_argument("--k" , default=1 , type=__a , help="k for the precision@k calculation" )
parser.add_argument(
"--evaluation_set" , default=__a , type=__a , required=__a , help="Path to a file containing evaluation samples" , )
parser.add_argument(
"--gold_data_path" , default=__a , type=__a , required=__a , help="Path to a tab-separated file with gold samples" , )
parser.add_argument(
"--gold_data_mode" , default="qa" , type=__a , choices=["qa", "ans"] , help=(
"Format of the gold data file"
"qa - a single line in the following format: question [tab] answer_list"
"ans - a single line of the gold file contains the expected answer string"
) , )
parser.add_argument(
"--predictions_path" , type=__a , default="predictions.txt" , help="Name of the predictions file, to be stored in the checkpoints directory" , )
parser.add_argument(
"--eval_all_checkpoints" , action="store_true" , help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number" , )
parser.add_argument(
"--eval_batch_size" , default=8 , type=__a , help="Batch size per GPU/CPU for evaluation." , )
parser.add_argument(
"--recalculate" , help="Recalculate predictions even if the prediction file exists" , action="store_true" , )
parser.add_argument(
"--num_beams" , default=4 , type=__a , help="Number of beams to be used when generating answers" , )
parser.add_argument("--min_length" , default=1 , type=__a , help="Min length of the generated answers" )
parser.add_argument("--max_length" , default=50 , type=__a , help="Max length of the generated answers" )
parser.add_argument(
"--print_predictions" , action="store_true" , help="If True, prints predictions while evaluating." , )
parser.add_argument(
"--print_docs" , action="store_true" , help="If True, prints docs retried while generating." , )
a__ : int = parser.parse_args()
a__ : Dict = torch.device("cuda" if torch.cuda.is_available() else "cpu" )
return args
def UpperCamelCase_ ( __a ) -> Optional[int]:
a__ : Tuple = {}
if args.model_type is None:
a__ : List[str] = infer_model_type(args.model_name_or_path )
assert args.model_type is not None
if args.model_type.startswith("rag" ):
a__ : int = RagTokenForGeneration if args.model_type == "rag_token" else RagSequenceForGeneration
a__ : Tuple = args.n_docs
if args.index_name is not None:
a__ : Any = args.index_name
if args.index_path is not None:
a__ : int = args.index_path
else:
a__ : Optional[Any] = BartForConditionalGeneration
a__ : Tuple = (
[f.path for f in os.scandir(args.model_name_or_path ) if f.is_dir()]
if args.eval_all_checkpoints
else [args.model_name_or_path]
)
logger.info("Evaluate the following checkpoints: %s" , __a )
a__ : Any = get_scores if args.eval_mode == "e2e" else get_precision_at_k
a__ : Union[str, Any] = evaluate_batch_eae if args.eval_mode == "e2e" else evaluate_batch_retrieval
for checkpoint in checkpoints:
if os.path.exists(args.predictions_path ) and (not args.recalculate):
logger.info("Calculating metrics based on an existing predictions file: {}".format(args.predictions_path ) )
score_fn(__a , args.predictions_path , args.gold_data_path )
continue
logger.info("***** Running evaluation for {} *****".format(__a ) )
logger.info(" Batch size = %d" , args.eval_batch_size )
logger.info(" Predictions will be stored under {}".format(args.predictions_path ) )
if args.model_type.startswith("rag" ):
a__ : str = RagRetriever.from_pretrained(__a , **__a )
a__ : Optional[int] = model_class.from_pretrained(__a , retriever=__a , **__a )
model.retriever.init_retrieval()
else:
a__ : Dict = model_class.from_pretrained(__a , **__a )
model.to(args.device )
with open(args.evaluation_set , "r" ) as eval_file, open(args.predictions_path , "w" ) as preds_file:
a__ : List[Any] = []
for line in tqdm(__a ):
questions.append(line.strip() )
if len(__a ) == args.eval_batch_size:
a__ : Union[str, Any] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) + "\n" )
preds_file.flush()
a__ : Any = []
if len(__a ) > 0:
a__ : List[str] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) )
preds_file.flush()
score_fn(__a , args.predictions_path , args.gold_data_path )
if __name__ == "__main__":
UpperCamelCase : List[Any] = get_args()
main(args)
| 37 | 1 |
def UpperCamelCase_ ( __a , __a ) -> int:
while b:
a__, a__ : int = b, a % b
return a
def UpperCamelCase_ ( __a , __a ) -> int:
return a if b == 0 else euclidean_gcd_recursive(__a , a % b )
def UpperCamelCase_ ( ) -> Dict:
print(f'''euclidean_gcd(3, 5) = {euclidean_gcd(3 , 5 )}''' )
print(f'''euclidean_gcd(5, 3) = {euclidean_gcd(5 , 3 )}''' )
print(f'''euclidean_gcd(1, 3) = {euclidean_gcd(1 , 3 )}''' )
print(f'''euclidean_gcd(3, 6) = {euclidean_gcd(3 , 6 )}''' )
print(f'''euclidean_gcd(6, 3) = {euclidean_gcd(6 , 3 )}''' )
print(f'''euclidean_gcd_recursive(3, 5) = {euclidean_gcd_recursive(3 , 5 )}''' )
print(f'''euclidean_gcd_recursive(5, 3) = {euclidean_gcd_recursive(5 , 3 )}''' )
print(f'''euclidean_gcd_recursive(1, 3) = {euclidean_gcd_recursive(1 , 3 )}''' )
print(f'''euclidean_gcd_recursive(3, 6) = {euclidean_gcd_recursive(3 , 6 )}''' )
print(f'''euclidean_gcd_recursive(6, 3) = {euclidean_gcd_recursive(6 , 3 )}''' )
if __name__ == "__main__":
main()
| 37 |
import logging
import os
from dataclasses import dataclass, field
from typing import Dict, Optional
import datasets
import numpy as np
import tensorflow as tf
from transformers import (
AutoConfig,
AutoTokenizer,
EvalPrediction,
HfArgumentParser,
PreTrainedTokenizer,
TFAutoModelForSequenceClassification,
TFTrainer,
TFTrainingArguments,
)
from transformers.utils import logging as hf_logging
hf_logging.set_verbosity_info()
hf_logging.enable_default_handler()
hf_logging.enable_explicit_format()
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a = None , ) -> str:
a__ : int = {}
if train_file is not None:
a__ : int = [train_file]
if eval_file is not None:
a__ : Union[str, Any] = [eval_file]
if test_file is not None:
a__ : str = [test_file]
a__ : Optional[Any] = datasets.load_dataset("csv" , data_files=__a )
a__ : List[Any] = list(ds[list(files.keys() )[0]].features.keys() )
a__ : str = features_name.pop(__a )
a__ : Dict = list(set(ds[list(files.keys() )[0]][label_name] ) )
a__ : str = {label: i for i, label in enumerate(__a )}
a__ : Tuple = tokenizer.model_input_names
a__ : List[str] = {}
if len(__a ) == 1:
for k in files.keys():
a__ : Optional[Any] = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
example[features_name[0]] , truncation=__a , max_length=__a , padding="max_length" ) , batched=__a , )
elif len(__a ) == 2:
for k in files.keys():
a__ : Dict = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
(example[features_name[0]], example[features_name[1]]) , truncation=__a , max_length=__a , padding="max_length" , ) , batched=__a , )
def gen_train():
for ex in transformed_ds[datasets.Split.TRAIN]:
a__ : str = {k: v for k, v in ex.items() if k in input_names}
a__ : str = labelaid[ex[label_name]]
yield (d, label)
def gen_val():
for ex in transformed_ds[datasets.Split.VALIDATION]:
a__ : Tuple = {k: v for k, v in ex.items() if k in input_names}
a__ : List[Any] = labelaid[ex[label_name]]
yield (d, label)
def gen_test():
for ex in transformed_ds[datasets.Split.TEST]:
a__ : List[Any] = {k: v for k, v in ex.items() if k in input_names}
a__ : Optional[int] = labelaid[ex[label_name]]
yield (d, label)
a__ : Optional[Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TRAIN in transformed_ds
else None
)
if train_ds is not None:
a__ : Optional[int] = train_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TRAIN] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.VALIDATION in transformed_ds
else None
)
if val_ds is not None:
a__ : Optional[Any] = val_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.VALIDATION] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TEST in transformed_ds
else None
)
if test_ds is not None:
a__ : Tuple = test_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TEST] ) ) )
return train_ds, val_ds, test_ds, labelaid
UpperCamelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(metadata={'help': 'Which column contains the label'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the training file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the development file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the test file'} )
_lowercase = field(
default=1_2_8 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
_lowercase = field(
default=A__ , metadata={'help': 'Overwrite the cached training and evaluation sets'} )
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
_lowercase = field(default=A__ , metadata={'help': 'Set this flag to use fast tokenization.'} )
# If you want to tweak more attributes on your tokenizer, you should do it in a distinct script,
# or just modify its tokenizer_config.json.
_lowercase = field(
default=A__ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
def UpperCamelCase_ ( ) -> Union[str, Any]:
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
a__ : str = HfArgumentParser((ModelArguments, DataTrainingArguments, TFTrainingArguments) )
a__, a__, a__ : str = parser.parse_args_into_dataclasses()
if (
os.path.exists(training_args.output_dir )
and os.listdir(training_args.output_dir )
and training_args.do_train
and not training_args.overwrite_output_dir
):
raise ValueError(
f'''Output directory ({training_args.output_dir}) already exists and is not empty. Use'''
" --overwrite_output_dir to overcome." )
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s" , datefmt="%m/%d/%Y %H:%M:%S" , level=logging.INFO , )
logger.info(
f'''n_replicas: {training_args.n_replicas}, distributed training: {bool(training_args.n_replicas > 1 )}, '''
f'''16-bits training: {training_args.fpaa}''' )
logger.info(f'''Training/evaluation parameters {training_args}''' )
# Load pretrained model and tokenizer
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
a__ : Union[str, Any] = AutoTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , )
a__, a__, a__, a__ : Optional[Any] = get_tfds(
train_file=data_args.train_file , eval_file=data_args.dev_file , test_file=data_args.test_file , tokenizer=__a , label_column_id=data_args.label_column_id , max_seq_length=data_args.max_seq_length , )
a__ : Optional[int] = AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=len(__a ) , labelaid=__a , idalabel={id: label for label, id in labelaid.items()} , finetuning_task="text-classification" , cache_dir=model_args.cache_dir , )
with training_args.strategy.scope():
a__ : Any = TFAutoModelForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_pt=bool(".bin" in model_args.model_name_or_path ) , config=__a , cache_dir=model_args.cache_dir , )
def compute_metrics(__a ) -> Dict:
a__ : Union[str, Any] = np.argmax(p.predictions , axis=1 )
return {"acc": (preds == p.label_ids).mean()}
# Initialize our Trainer
a__ : Dict = TFTrainer(
model=__a , args=__a , train_dataset=__a , eval_dataset=__a , compute_metrics=__a , )
# Training
if training_args.do_train:
trainer.train()
trainer.save_model()
tokenizer.save_pretrained(training_args.output_dir )
# Evaluation
a__ : Optional[Any] = {}
if training_args.do_eval:
logger.info("*** Evaluate ***" )
a__ : Dict = trainer.evaluate()
a__ : int = os.path.join(training_args.output_dir , "eval_results.txt" )
with open(__a , "w" ) as writer:
logger.info("***** Eval results *****" )
for key, value in result.items():
logger.info(f''' {key} = {value}''' )
writer.write(f'''{key} = {value}\n''' )
results.update(__a )
return results
if __name__ == "__main__":
main()
| 37 | 1 |
import unittest
import numpy as np
from transformers import RobertaPreLayerNormConfig, is_flax_available
from transformers.testing_utils import require_flax, slow
from ...test_modeling_flax_common import FlaxModelTesterMixin, floats_tensor, ids_tensor, random_attention_mask
if is_flax_available():
import jax.numpy as jnp
from transformers.models.roberta_prelayernorm.modeling_flax_roberta_prelayernorm import (
FlaxRobertaPreLayerNormForCausalLM,
FlaxRobertaPreLayerNormForMaskedLM,
FlaxRobertaPreLayerNormForMultipleChoice,
FlaxRobertaPreLayerNormForQuestionAnswering,
FlaxRobertaPreLayerNormForSequenceClassification,
FlaxRobertaPreLayerNormForTokenClassification,
FlaxRobertaPreLayerNormModel,
)
class A__ ( unittest.TestCase ):
"""simple docstring"""
def __init__( self : Tuple , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Union[str, Any]=13 , lowerCamelCase__ : str=7 , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : Optional[int]=True , lowerCamelCase__ : Union[str, Any]=True , lowerCamelCase__ : List[Any]=True , lowerCamelCase__ : Tuple=99 , lowerCamelCase__ : int=32 , lowerCamelCase__ : List[Any]=5 , lowerCamelCase__ : str=4 , lowerCamelCase__ : str=37 , lowerCamelCase__ : Dict="gelu" , lowerCamelCase__ : Tuple=0.1 , lowerCamelCase__ : Tuple=0.1 , lowerCamelCase__ : Tuple=512 , lowerCamelCase__ : Optional[int]=16 , lowerCamelCase__ : List[str]=2 , lowerCamelCase__ : Union[str, Any]=0.02 , lowerCamelCase__ : List[Any]=4 , ):
a__ : Optional[Any] = parent
a__ : List[str] = batch_size
a__ : str = seq_length
a__ : Optional[int] = is_training
a__ : Optional[Any] = use_attention_mask
a__ : str = use_token_type_ids
a__ : Optional[Any] = use_labels
a__ : str = vocab_size
a__ : Optional[Any] = hidden_size
a__ : str = num_hidden_layers
a__ : int = num_attention_heads
a__ : Tuple = intermediate_size
a__ : List[Any] = hidden_act
a__ : Optional[int] = hidden_dropout_prob
a__ : Any = attention_probs_dropout_prob
a__ : str = max_position_embeddings
a__ : Optional[int] = type_vocab_size
a__ : Tuple = type_sequence_label_size
a__ : Optional[int] = initializer_range
a__ : List[Any] = num_choices
def _UpperCamelCase( self : Optional[int] ):
a__ : Any = ids_tensor([self.batch_size, self.seq_length] , self.vocab_size )
a__ : int = None
if self.use_attention_mask:
a__ : str = random_attention_mask([self.batch_size, self.seq_length] )
a__ : Dict = None
if self.use_token_type_ids:
a__ : Optional[Any] = ids_tensor([self.batch_size, self.seq_length] , self.type_vocab_size )
a__ : int = RobertaPreLayerNormConfig(
vocab_size=self.vocab_size , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , max_position_embeddings=self.max_position_embeddings , type_vocab_size=self.type_vocab_size , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , )
return config, input_ids, token_type_ids, attention_mask
def _UpperCamelCase( self : Tuple ):
a__ : str = self.prepare_config_and_inputs()
a__, a__, a__, a__ : Optional[Any] = config_and_inputs
a__ : Optional[int] = {"input_ids": input_ids, "token_type_ids": token_type_ids, "attention_mask": attention_mask}
return config, inputs_dict
def _UpperCamelCase( self : str ):
a__ : Optional[Any] = self.prepare_config_and_inputs()
a__, a__, a__, a__ : Union[str, Any] = config_and_inputs
a__ : Optional[int] = True
a__ : Dict = floats_tensor([self.batch_size, self.seq_length, self.hidden_size] )
a__ : Union[str, Any] = ids_tensor([self.batch_size, self.seq_length] , vocab_size=2 )
return (
config,
input_ids,
token_type_ids,
encoder_hidden_states,
encoder_attention_mask,
)
@require_flax
# Copied from tests.models.roberta.test_modelling_flax_roberta.FlaxRobertaPreLayerNormModelTest with ROBERTA->ROBERTA_PRELAYERNORM,Roberta->RobertaPreLayerNorm,roberta-base->andreasmadsen/efficient_mlm_m0.40
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = True
_lowercase = (
(
FlaxRobertaPreLayerNormModel,
FlaxRobertaPreLayerNormForCausalLM,
FlaxRobertaPreLayerNormForMaskedLM,
FlaxRobertaPreLayerNormForSequenceClassification,
FlaxRobertaPreLayerNormForTokenClassification,
FlaxRobertaPreLayerNormForMultipleChoice,
FlaxRobertaPreLayerNormForQuestionAnswering,
)
if is_flax_available()
else ()
)
def _UpperCamelCase( self : List[str] ):
a__ : Dict = FlaxRobertaPreLayerNormModelTester(self )
@slow
def _UpperCamelCase( self : Dict ):
for model_class_name in self.all_model_classes:
a__ : List[str] = model_class_name.from_pretrained("andreasmadsen/efficient_mlm_m0.40" , from_pt=lowerCamelCase__ )
a__ : Optional[int] = model(np.ones((1, 1) ) )
self.assertIsNotNone(lowerCamelCase__ )
@require_flax
class A__ ( unittest.TestCase ):
"""simple docstring"""
@slow
def _UpperCamelCase( self : Dict ):
a__ : Any = FlaxRobertaPreLayerNormForMaskedLM.from_pretrained("andreasmadsen/efficient_mlm_m0.40" , from_pt=lowerCamelCase__ )
a__ : str = np.array([[0, 31_414, 232, 328, 740, 1_140, 12_695, 69, 46_078, 1_588, 2]] , dtype=jnp.intaa )
a__ : int = model(lowerCamelCase__ )[0]
a__ : int = [1, 11, 50_265]
self.assertEqual(list(output.shape ) , lowerCamelCase__ )
# compare the actual values for a slice.
a__ : str = np.array(
[[[40.4880, 18.0199, -5.2367], [-1.8877, -4.0885, 10.7085], [-2.2613, -5.6110, 7.2665]]] , dtype=np.floataa )
self.assertTrue(np.allclose(output[:, :3, :3] , lowerCamelCase__ , atol=1E-4 ) )
@slow
def _UpperCamelCase( self : Optional[Any] ):
a__ : int = FlaxRobertaPreLayerNormModel.from_pretrained("andreasmadsen/efficient_mlm_m0.40" , from_pt=lowerCamelCase__ )
a__ : Any = np.array([[0, 31_414, 232, 328, 740, 1_140, 12_695, 69, 46_078, 1_588, 2]] , dtype=jnp.intaa )
a__ : str = model(lowerCamelCase__ )[0]
# compare the actual values for a slice.
a__ : Optional[Any] = np.array(
[[[0.0208, -0.0356, 0.0237], [-0.1569, -0.0411, -0.2626], [0.1879, 0.0125, -0.0089]]] , dtype=np.floataa )
self.assertTrue(np.allclose(output[:, :3, :3] , lowerCamelCase__ , atol=1E-4 ) )
| 37 |
import argparse
import collections
import json
import os
import re
import string
import sys
import numpy as np
UpperCamelCase : List[str] = re.compile(r"""\b(a|an|the)\b""", re.UNICODE)
UpperCamelCase : Union[str, Any] = None
def UpperCamelCase_ ( ) -> List[str]:
a__ : List[Any] = argparse.ArgumentParser("Official evaluation script for SQuAD version 2.0." )
parser.add_argument("data_file" , metavar="data.json" , help="Input data JSON file." )
parser.add_argument("pred_file" , metavar="pred.json" , help="Model predictions." )
parser.add_argument(
"--out-file" , "-o" , metavar="eval.json" , help="Write accuracy metrics to file (default is stdout)." )
parser.add_argument(
"--na-prob-file" , "-n" , metavar="na_prob.json" , help="Model estimates of probability of no answer." )
parser.add_argument(
"--na-prob-thresh" , "-t" , type=__a , default=1.0 , help="Predict \"\" if no-answer probability exceeds this (default = 1.0)." , )
parser.add_argument(
"--out-image-dir" , "-p" , metavar="out_images" , default=__a , help="Save precision-recall curves to directory." )
parser.add_argument("--verbose" , "-v" , action="store_true" )
if len(sys.argv ) == 1:
parser.print_help()
sys.exit(1 )
return parser.parse_args()
def UpperCamelCase_ ( __a ) -> str:
a__ : Optional[Any] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : Dict = bool(qa["answers"]["text"] )
return qid_to_has_ans
def UpperCamelCase_ ( __a ) -> List[Any]:
def remove_articles(__a ):
return ARTICLES_REGEX.sub(" " , __a )
def white_space_fix(__a ):
return " ".join(text.split() )
def remove_punc(__a ):
a__ : Union[str, Any] = set(string.punctuation )
return "".join(ch for ch in text if ch not in exclude )
def lower(__a ):
return text.lower()
return white_space_fix(remove_articles(remove_punc(lower(__a ) ) ) )
def UpperCamelCase_ ( __a ) -> Dict:
if not s:
return []
return normalize_answer(__a ).split()
def UpperCamelCase_ ( __a , __a ) -> str:
return int(normalize_answer(__a ) == normalize_answer(__a ) )
def UpperCamelCase_ ( __a , __a ) -> Dict:
a__ : int = get_tokens(__a )
a__ : Optional[Any] = get_tokens(__a )
a__ : Any = collections.Counter(__a ) & collections.Counter(__a )
a__ : Dict = sum(common.values() )
if len(__a ) == 0 or len(__a ) == 0:
# If either is no-answer, then F1 is 1 if they agree, 0 otherwise
return int(gold_toks == pred_toks )
if num_same == 0:
return 0
a__ : Tuple = 1.0 * num_same / len(__a )
a__ : str = 1.0 * num_same / len(__a )
a__ : str = (2 * precision * recall) / (precision + recall)
return fa
def UpperCamelCase_ ( __a , __a ) -> int:
a__ : List[str] = {}
a__ : Optional[int] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : List[Any] = qa["id"]
a__ : Dict = [t for t in qa["answers"]["text"] if normalize_answer(__a )]
if not gold_answers:
# For unanswerable questions, only correct answer is empty string
a__ : Tuple = [""]
if qid not in preds:
print(f'''Missing prediction for {qid}''' )
continue
a__ : Tuple = preds[qid]
# Take max over all gold answers
a__ : Optional[int] = max(compute_exact(__a , __a ) for a in gold_answers )
a__ : str = max(compute_fa(__a , __a ) for a in gold_answers )
return exact_scores, fa_scores
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
a__ : Optional[Any] = {}
for qid, s in scores.items():
a__ : Dict = na_probs[qid] > na_prob_thresh
if pred_na:
a__ : Dict = float(not qid_to_has_ans[qid] )
else:
a__ : Optional[Any] = s
return new_scores
def UpperCamelCase_ ( __a , __a , __a=None ) -> Tuple:
if not qid_list:
a__ : Union[str, Any] = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores.values() ) / total),
("f1", 100.0 * sum(fa_scores.values() ) / total),
("total", total),
] )
else:
a__ : int = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores[k] for k in qid_list ) / total),
("f1", 100.0 * sum(fa_scores[k] for k in qid_list ) / total),
("total", total),
] )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
for k in new_eval:
a__ : Optional[Any] = new_eval[k]
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
plt.step(__a , __a , color="b" , alpha=0.2 , where="post" )
plt.fill_between(__a , __a , step="post" , alpha=0.2 , color="b" )
plt.xlabel("Recall" )
plt.ylabel("Precision" )
plt.xlim([0.0, 1.05] )
plt.ylim([0.0, 1.05] )
plt.title(__a )
plt.savefig(__a )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a , __a=None , __a=None ) -> Dict:
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
a__ : Any = 0.0
a__ : Optional[int] = 1.0
a__ : Optional[int] = 0.0
a__ : Any = [1.0]
a__ : Tuple = [0.0]
a__ : List[str] = 0.0
for i, qid in enumerate(__a ):
if qid_to_has_ans[qid]:
true_pos += scores[qid]
a__ : Any = true_pos / float(i + 1 )
a__ : int = true_pos / float(__a )
if i == len(__a ) - 1 or na_probs[qid] != na_probs[qid_list[i + 1]]:
# i.e., if we can put a threshold after this point
avg_prec += cur_p * (cur_r - recalls[-1])
precisions.append(__a )
recalls.append(__a )
if out_image:
plot_pr_curve(__a , __a , __a , __a )
return {"ap": 100.0 * avg_prec}
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> str:
if out_image_dir and not os.path.exists(__a ):
os.makedirs(__a )
a__ : Optional[int] = sum(1 for v in qid_to_has_ans.values() if v )
if num_true_pos == 0:
return
a__ : Optional[int] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_exact.png" ) , title="Precision-Recall curve for Exact Match score" , )
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_f1.png" ) , title="Precision-Recall curve for F1 score" , )
a__ : str = {k: float(__a ) for k, v in qid_to_has_ans.items()}
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_oracle.png" ) , title="Oracle Precision-Recall curve (binary task of HasAns vs. NoAns)" , )
merge_eval(__a , __a , "pr_exact" )
merge_eval(__a , __a , "pr_f1" )
merge_eval(__a , __a , "pr_oracle" )
def UpperCamelCase_ ( __a , __a , __a , __a ) -> str:
if not qid_list:
return
a__ : Optional[Any] = [na_probs[k] for k in qid_list]
a__ : str = np.ones_like(__a ) / float(len(__a ) )
plt.hist(__a , weights=__a , bins=20 , range=(0.0, 1.0) )
plt.xlabel("Model probability of no-answer" )
plt.ylabel("Proportion of dataset" )
plt.title(f'''Histogram of no-answer probability: {name}''' )
plt.savefig(os.path.join(__a , f'''na_prob_hist_{name}.png''' ) )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[Any]:
a__ : str = sum(1 for k in qid_to_has_ans if not qid_to_has_ans[k] )
a__ : Optional[Any] = num_no_ans
a__ : Dict = cur_score
a__ : Any = 0.0
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
for i, qid in enumerate(__a ):
if qid not in scores:
continue
if qid_to_has_ans[qid]:
a__ : Optional[int] = scores[qid]
else:
if preds[qid]:
a__ : str = -1
else:
a__ : Union[str, Any] = 0
cur_score += diff
if cur_score > best_score:
a__ : Any = cur_score
a__ : Dict = na_probs[qid]
return 100.0 * best_score / len(__a ), best_thresh
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> Any:
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__ : Any = best_exact
a__ : Any = exact_thresh
a__ : List[Any] = best_fa
a__ : Optional[int] = fa_thresh
def UpperCamelCase_ ( ) -> Tuple:
with open(OPTS.data_file ) as f:
a__ : List[Any] = json.load(__a )
a__ : Any = dataset_json["data"]
with open(OPTS.pred_file ) as f:
a__ : int = json.load(__a )
if OPTS.na_prob_file:
with open(OPTS.na_prob_file ) as f:
a__ : List[str] = json.load(__a )
else:
a__ : Optional[int] = {k: 0.0 for k in preds}
a__ : Optional[Any] = make_qid_to_has_ans(__a ) # maps qid to True/False
a__ : List[Any] = [k for k, v in qid_to_has_ans.items() if v]
a__ : Union[str, Any] = [k for k, v in qid_to_has_ans.items() if not v]
a__, a__ : str = get_raw_scores(__a , __a )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : Tuple = make_eval_dict(__a , __a )
if has_ans_qids:
a__ : str = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "HasAns" )
if no_ans_qids:
a__ : List[Any] = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "NoAns" )
if OPTS.na_prob_file:
find_all_best_thresh(__a , __a , __a , __a , __a , __a )
if OPTS.na_prob_file and OPTS.out_image_dir:
run_precision_recall_analysis(__a , __a , __a , __a , __a , OPTS.out_image_dir )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "hasAns" )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "noAns" )
if OPTS.out_file:
with open(OPTS.out_file , "w" ) as f:
json.dump(__a , __a )
else:
print(json.dumps(__a , indent=2 ) )
if __name__ == "__main__":
UpperCamelCase : Any = parse_args()
if OPTS.out_image_dir:
import matplotlib
matplotlib.use("""Agg""")
import matplotlib.pyplot as plt
main()
| 37 | 1 |
import re
from filelock import FileLock
try:
import nltk
UpperCamelCase : Union[str, Any] = True
except (ImportError, ModuleNotFoundError):
UpperCamelCase : Any = False
if NLTK_AVAILABLE:
with FileLock(""".lock""") as lock:
nltk.download("""punkt""", quiet=True)
def UpperCamelCase_ ( __a ) -> str:
re.sub("<n>" , "" , __a ) # remove pegasus newline char
assert NLTK_AVAILABLE, "nltk must be installed to separate newlines between sentences. (pip install nltk)"
return "\n".join(nltk.sent_tokenize(__a ) )
| 37 |
import json
import os
import unittest
from transformers import CLIPTokenizer, CLIPTokenizerFast
from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_ftfy, require_tokenizers
from ...test_tokenization_common import TokenizerTesterMixin
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = CLIPTokenizer
_lowercase = CLIPTokenizerFast
_lowercase = True
_lowercase = {}
_lowercase = False
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# fmt: off
a__ : Any = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "lo", "l</w>", "w</w>", "r</w>", "t</w>", "low</w>", "er</w>", "lowest</w>", "newer</w>", "wider", "<unk>", "<|startoftext|>", "<|endoftext|>"]
# fmt: on
a__ : Optional[Any] = dict(zip(lowerCamelCase__ , range(len(lowerCamelCase__ ) ) ) )
a__ : Optional[Any] = ["#version: 0.2", "l o", "lo w</w>", "e r</w>"]
a__ : Optional[Any] = {"unk_token": "<unk>"}
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["vocab_file"] )
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["merges_file"] )
with open(self.vocab_file , "w" , encoding="utf-8" ) as fp:
fp.write(json.dumps(lowerCamelCase__ ) + "\n" )
with open(self.merges_file , "w" , encoding="utf-8" ) as fp:
fp.write("\n".join(lowerCamelCase__ ) )
def _UpperCamelCase( self : Dict , **lowerCamelCase__ : int ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizer.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , **lowerCamelCase__ : Optional[int] ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizerFast.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[Any] ):
a__ : int = "lower newer"
a__ : Optional[int] = "lower newer"
return input_text, output_text
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = CLIPTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
a__ : int = "lower newer"
a__ : List[str] = ["lo", "w", "er</w>", "n", "e", "w", "er</w>"]
a__ : Union[str, Any] = tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : int = tokens + [tokenizer.unk_token]
a__ : Union[str, Any] = [10, 2, 16, 9, 3, 2, 16, 20]
self.assertListEqual(tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , lowerCamelCase__ )
@require_ftfy
def _UpperCamelCase( self : Optional[Any] ):
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : List[str] = self.tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : Any = self.rust_tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : int = "A\n'll 11p223RF☆ho!!to?'d'd''d of a cat to-$''d."
a__ : Optional[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : Dict = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on an example containing a character (Latin Small Letter A
# with Tilde) encoded in 2 different ways
a__ : Optional[Any] = "xa\u0303y" + " " + "x\xe3y"
a__ : Optional[int] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of space type
a__ : str = [
"\u0009", # (horizontal tab, '\t')
"\u000B", # (vertical tab)
"\u000C", # (form feed)
"\u0020", # (space, ' ')
"\u200E", # (left-to-right mark):w
"\u200F", # (right-to-left mark)
]
for unicode_seq in spaces_unicodes:
a__ : Any = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of line break type
a__ : Union[str, Any] = [
"\u000A", # (line feed, '\n')
"\r\n", # (carriage return and line feed, '\r\n')
"\u000D", # (carriage return, '\r')
"\r", # (carriage return, '\r')
"\u000D", # (carriage return, '\r')
"\u2028", # (line separator)
"\u2029", # (paragraph separator)
# "\u0085", # (next line)
]
# The tokenization is not identical for the character "\u0085" (next line). The slow version using ftfy transforms
# it into the Horizontal Ellipsis character "…" ("\u2026") while the fast version transforms it into a
# space (and thus into an empty list).
for unicode_seq in line_break_unicodes:
a__ : List[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
# Test which aims to verify that the offsets are well adapted to the argument `add_prefix_space`
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : str = "hello" # `hello` is a token in the vocabulary of `pretrained_name`
a__ : Tuple = f'''{text_of_1_token} {text_of_1_token}'''
a__ : Optional[int] = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Union[str, Any] = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (0, len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (len(lowerCamelCase__ ) + 1, len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
a__ : Optional[Any] = f''' {text}'''
a__ : str = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Dict = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (1, 1 + len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (1 + len(lowerCamelCase__ ) + 1, 1 + len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
def _UpperCamelCase( self : int ):
# Test related to the breaking change introduced in transformers v4.17.0
# We need to check that an error in raised when the user try to load a previous version of the tokenizer.
with self.assertRaises(lowerCamelCase__ ) as context:
self.rust_tokenizer_class.from_pretrained("robot-test/old-clip-tokenizer" )
self.assertTrue(
context.exception.args[0].startswith(
"The `backend_tokenizer` provided does not match the expected format." ) )
@require_ftfy
def _UpperCamelCase( self : int ):
super().test_tokenization_python_rust_equals()
def _UpperCamelCase( self : str ):
# CLIP always lower cases letters
pass
| 37 | 1 |
import os
import shutil
from pathlib import Path
from typing import Optional, Union
import numpy as np
from huggingface_hub import hf_hub_download
from ..utils import ONNX_EXTERNAL_WEIGHTS_NAME, ONNX_WEIGHTS_NAME, is_onnx_available, logging
if is_onnx_available():
import onnxruntime as ort
UpperCamelCase : Any = logging.get_logger(__name__)
UpperCamelCase : Optional[Any] = {
"""tensor(bool)""": np.bool_,
"""tensor(int8)""": np.inta,
"""tensor(uint8)""": np.uinta,
"""tensor(int16)""": np.intaa,
"""tensor(uint16)""": np.uintaa,
"""tensor(int32)""": np.intaa,
"""tensor(uint32)""": np.uintaa,
"""tensor(int64)""": np.intaa,
"""tensor(uint64)""": np.uintaa,
"""tensor(float16)""": np.floataa,
"""tensor(float)""": np.floataa,
"""tensor(double)""": np.floataa,
}
class A__ :
"""simple docstring"""
def __init__( self : str , lowerCamelCase__ : int=None , **lowerCamelCase__ : Tuple ):
logger.info("`diffusers.OnnxRuntimeModel` is experimental and might change in the future." )
a__ : Optional[int] = model
a__ : str = kwargs.get("model_save_dir" , lowerCamelCase__ )
a__ : List[str] = kwargs.get("latest_model_name" , lowerCamelCase__ )
def __call__( self : Tuple , **lowerCamelCase__ : Tuple ):
a__ : Union[str, Any] = {k: np.array(lowerCamelCase__ ) for k, v in kwargs.items()}
return self.model.run(lowerCamelCase__ , lowerCamelCase__ )
@staticmethod
def _UpperCamelCase( lowerCamelCase__ : Union[str, Path] , lowerCamelCase__ : List[Any]=None , lowerCamelCase__ : Optional[Any]=None ):
if provider is None:
logger.info("No onnxruntime provider specified, using CPUExecutionProvider" )
a__ : List[Any] = "CPUExecutionProvider"
return ort.InferenceSession(lowerCamelCase__ , providers=[provider] , sess_options=lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Union[str, Path] , lowerCamelCase__ : Optional[str] = None , **lowerCamelCase__ : Tuple ):
a__ : Union[str, Any] = file_name if file_name is not None else ONNX_WEIGHTS_NAME
a__ : Union[str, Any] = self.model_save_dir.joinpath(self.latest_model_name )
a__ : Optional[int] = Path(lowerCamelCase__ ).joinpath(lowerCamelCase__ )
try:
shutil.copyfile(lowerCamelCase__ , lowerCamelCase__ )
except shutil.SameFileError:
pass
# copy external weights (for models >2GB)
a__ : Optional[int] = self.model_save_dir.joinpath(lowerCamelCase__ )
if src_path.exists():
a__ : Union[str, Any] = Path(lowerCamelCase__ ).joinpath(lowerCamelCase__ )
try:
shutil.copyfile(lowerCamelCase__ , lowerCamelCase__ )
except shutil.SameFileError:
pass
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Union[str, os.PathLike] , **lowerCamelCase__ : Optional[Any] , ):
if os.path.isfile(lowerCamelCase__ ):
logger.error(f'''Provided path ({save_directory}) should be a directory, not a file''' )
return
os.makedirs(lowerCamelCase__ , exist_ok=lowerCamelCase__ )
# saving model weights/files
self._save_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
@classmethod
def _UpperCamelCase( cls : List[Any] , lowerCamelCase__ : Union[str, Path] , lowerCamelCase__ : Optional[Union[bool, str, None]] = None , lowerCamelCase__ : Optional[Union[str, None]] = None , lowerCamelCase__ : bool = False , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional["ort.SessionOptions"] = None , **lowerCamelCase__ : Union[str, Any] , ):
a__ : int = file_name if file_name is not None else ONNX_WEIGHTS_NAME
# load model from local directory
if os.path.isdir(lowerCamelCase__ ):
a__ : Optional[Any] = OnnxRuntimeModel.load_model(
os.path.join(lowerCamelCase__ , lowerCamelCase__ ) , provider=lowerCamelCase__ , sess_options=lowerCamelCase__ )
a__ : List[Any] = Path(lowerCamelCase__ )
# load model from hub
else:
# download model
a__ : Any = hf_hub_download(
repo_id=lowerCamelCase__ , filename=lowerCamelCase__ , use_auth_token=lowerCamelCase__ , revision=lowerCamelCase__ , cache_dir=lowerCamelCase__ , force_download=lowerCamelCase__ , )
a__ : Optional[int] = Path(lowerCamelCase__ ).parent
a__ : Union[str, Any] = Path(lowerCamelCase__ ).name
a__ : Union[str, Any] = OnnxRuntimeModel.load_model(lowerCamelCase__ , provider=lowerCamelCase__ , sess_options=lowerCamelCase__ )
return cls(model=lowerCamelCase__ , **lowerCamelCase__ )
@classmethod
def _UpperCamelCase( cls : Dict , lowerCamelCase__ : Union[str, Path] , lowerCamelCase__ : bool = True , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[str] = None , **lowerCamelCase__ : List[str] , ):
a__ : List[str] = None
if len(str(lowerCamelCase__ ).split("@" ) ) == 2:
a__, a__ : Optional[Any] = model_id.split("@" )
return cls._from_pretrained(
model_id=lowerCamelCase__ , revision=lowerCamelCase__ , cache_dir=lowerCamelCase__ , force_download=lowerCamelCase__ , use_auth_token=lowerCamelCase__ , **lowerCamelCase__ , )
| 37 |
import os
import re
import shutil
from argparse import ArgumentParser, Namespace
from datasets.commands import BaseDatasetsCLICommand
from datasets.utils.logging import get_logger
UpperCamelCase : Dict = """<<<<<<< This should probably be modified because it mentions: """
UpperCamelCase : List[Any] = """=======
>>>>>>>
"""
UpperCamelCase : Optional[Any] = [
"""TextEncoderConfig""",
"""ByteTextEncoder""",
"""SubwordTextEncoder""",
"""encoder_config""",
"""maybe_build_from_corpus""",
"""manual_dir""",
]
UpperCamelCase : Any = [
# (pattern, replacement)
# Order is important here for some replacements
(r"""tfds\.core""", r"""datasets"""),
(r"""tf\.io\.gfile\.GFile""", r"""open"""),
(r"""tf\.([\w\d]+)""", r"""datasets.Value('\1')"""),
(r"""tfds\.features\.Text\(\)""", r"""datasets.Value('string')"""),
(r"""tfds\.features\.Text\(""", r"""datasets.Value('string'),"""),
(r"""features\s*=\s*tfds.features.FeaturesDict\(""", r"""features=datasets.Features("""),
(r"""tfds\.features\.FeaturesDict\(""", r"""dict("""),
(r"""The TensorFlow Datasets Authors""", r"""The TensorFlow Datasets Authors and the HuggingFace Datasets Authors"""),
(r"""tfds\.""", r"""datasets."""),
(r"""dl_manager\.manual_dir""", r"""self.config.data_dir"""),
(r"""self\.builder_config""", r"""self.config"""),
]
def UpperCamelCase_ ( __a ) -> Optional[Any]:
return ConvertCommand(args.tfds_path , args.datasets_directory )
class A__ ( A__ ):
"""simple docstring"""
@staticmethod
def _UpperCamelCase( lowerCamelCase__ : ArgumentParser ):
a__ : List[str] = parser.add_parser(
"convert" , help="Convert a TensorFlow Datasets dataset to a HuggingFace Datasets dataset." , )
train_parser.add_argument(
"--tfds_path" , type=lowerCamelCase__ , required=lowerCamelCase__ , help="Path to a TensorFlow Datasets folder to convert or a single tfds file to convert." , )
train_parser.add_argument(
"--datasets_directory" , type=lowerCamelCase__ , required=lowerCamelCase__ , help="Path to the HuggingFace Datasets folder." )
train_parser.set_defaults(func=lowerCamelCase__ )
def __init__( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str , *lowerCamelCase__ : Tuple ):
a__ : str = get_logger("datasets-cli/converting" )
a__ : Optional[Any] = tfds_path
a__ : Optional[int] = datasets_directory
def _UpperCamelCase( self : int ):
if os.path.isdir(self._tfds_path ):
a__ : List[str] = os.path.abspath(self._tfds_path )
elif os.path.isfile(self._tfds_path ):
a__ : Any = os.path.dirname(self._tfds_path )
else:
raise ValueError("--tfds_path is neither a directory nor a file. Please check path." )
a__ : Dict = os.path.abspath(self._datasets_directory )
self._logger.info(f'''Converting datasets from {abs_tfds_path} to {abs_datasets_path}''' )
a__ : Tuple = []
a__ : str = []
a__ : List[Any] = {}
if os.path.isdir(self._tfds_path ):
a__ : List[str] = os.listdir(lowerCamelCase__ )
else:
a__ : Union[str, Any] = [os.path.basename(self._tfds_path )]
for f_name in file_names:
self._logger.info(f'''Looking at file {f_name}''' )
a__ : Any = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
a__ : Dict = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
if not os.path.isfile(lowerCamelCase__ ) or "__init__" in f_name or "_test" in f_name or ".py" not in f_name:
self._logger.info("Skipping file" )
continue
with open(lowerCamelCase__ , encoding="utf-8" ) as f:
a__ : List[Any] = f.readlines()
a__ : Union[str, Any] = []
a__ : Union[str, Any] = False
a__ : Union[str, Any] = False
a__ : Dict = []
for line in lines:
a__ : Optional[Any] = line
# Convert imports
if "import tensorflow.compat.v2 as tf" in out_line:
continue
elif "@tfds.core" in out_line:
continue
elif "builder=self" in out_line:
continue
elif "import tensorflow_datasets.public_api as tfds" in out_line:
a__ : List[Any] = "import datasets\n"
elif "import tensorflow" in out_line:
# order is important here
a__ : List[str] = ""
continue
elif "from absl import logging" in out_line:
a__ : Dict = "from datasets import logging\n"
elif "getLogger" in out_line:
a__ : List[Any] = out_line.replace("getLogger" , "get_logger" )
elif any(expression in out_line for expression in TO_HIGHLIGHT ):
a__ : List[str] = True
a__ : Dict = list(filter(lambda lowerCamelCase__ : e in out_line , lowerCamelCase__ ) )
out_lines.append(HIGHLIGHT_MESSAGE_PRE + str(lowerCamelCase__ ) + "\n" )
out_lines.append(lowerCamelCase__ )
out_lines.append(lowerCamelCase__ )
continue
else:
for pattern, replacement in TO_CONVERT:
a__ : Tuple = re.sub(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
# Take care of saving utilities (to later move them together with main script)
if "tensorflow_datasets" in out_line:
a__ : Optional[int] = re.match(r"from\stensorflow_datasets.*import\s([^\.\r\n]+)" , lowerCamelCase__ )
tfds_imports.extend(imp.strip() for imp in match.group(1 ).split("," ) )
a__ : Optional[Any] = "from . import " + match.group(1 )
# Check we have not forget anything
if "tf." in out_line or "tfds." in out_line or "tensorflow_datasets" in out_line:
raise ValueError(f'''Error converting {out_line.strip()}''' )
if "GeneratorBasedBuilder" in out_line or "BeamBasedBuilder" in out_line:
a__ : Optional[int] = True
out_lines.append(lowerCamelCase__ )
if is_builder or "wmt" in f_name:
# We create a new directory for each dataset
a__ : Dict = f_name.replace(".py" , "" )
a__ : Optional[int] = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
a__ : Any = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
os.makedirs(lowerCamelCase__ , exist_ok=lowerCamelCase__ )
self._logger.info(f'''Adding directory {output_dir}''' )
imports_to_builder_map.update({imp: output_dir for imp in tfds_imports} )
else:
# Utilities will be moved at the end
utils_files.append(lowerCamelCase__ )
if needs_manual_update:
with_manual_update.append(lowerCamelCase__ )
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as f:
f.writelines(lowerCamelCase__ )
self._logger.info(f'''Converted in {output_file}''' )
for utils_file in utils_files:
try:
a__ : Any = os.path.basename(lowerCamelCase__ )
a__ : Optional[int] = imports_to_builder_map[f_name.replace(".py" , "" )]
self._logger.info(f'''Moving {dest_folder} to {utils_file}''' )
shutil.copy(lowerCamelCase__ , lowerCamelCase__ )
except KeyError:
self._logger.error(f'''Cannot find destination folder for {utils_file}. Please copy manually.''' )
if with_manual_update:
for file_path in with_manual_update:
self._logger.warning(
f'''You need to manually update file {file_path} to remove configurations using \'TextEncoderConfig\'.''' )
| 37 | 1 |
from __future__ import annotations
import queue
class A__ :
"""simple docstring"""
def __init__( self : Dict , lowerCamelCase__ : Optional[Any] ):
a__ : Any = data
a__ : str = None
a__ : str = None
def UpperCamelCase_ ( ) -> TreeNode:
print("\n********Press N to stop entering at any point of time********\n" )
a__ : Union[str, Any] = input("Enter the value of the root node: " ).strip().lower()
a__ : queue.Queue = queue.Queue()
a__ : List[Any] = TreeNode(int(__a ) )
q.put(__a )
while not q.empty():
a__ : Dict = q.get()
a__ : str = f'''Enter the left node of {node_found.data}: '''
a__ : Any = input(__a ).strip().lower() or "n"
if check == "n":
return tree_node
a__ : Any = TreeNode(int(__a ) )
a__ : str = left_node
q.put(__a )
a__ : Optional[Any] = f'''Enter the right node of {node_found.data}: '''
a__ : Any = input(__a ).strip().lower() or "n"
if check == "n":
return tree_node
a__ : List[Any] = TreeNode(int(__a ) )
a__ : Tuple = right_node
q.put(__a )
raise
def UpperCamelCase_ ( __a ) -> None:
if not isinstance(__a , __a ) or not node:
return
print(node.data , end="," )
pre_order(node.left )
pre_order(node.right )
def UpperCamelCase_ ( __a ) -> None:
if not isinstance(__a , __a ) or not node:
return
in_order(node.left )
print(node.data , end="," )
in_order(node.right )
def UpperCamelCase_ ( __a ) -> None:
if not isinstance(__a , __a ) or not node:
return
post_order(node.left )
post_order(node.right )
print(node.data , end="," )
def UpperCamelCase_ ( __a ) -> None:
if not isinstance(__a , __a ) or not node:
return
a__ : queue.Queue = queue.Queue()
q.put(__a )
while not q.empty():
a__ : Union[str, Any] = q.get()
print(node_dequeued.data , end="," )
if node_dequeued.left:
q.put(node_dequeued.left )
if node_dequeued.right:
q.put(node_dequeued.right )
def UpperCamelCase_ ( __a ) -> None:
if not isinstance(__a , __a ) or not node:
return
a__ : queue.Queue = queue.Queue()
q.put(__a )
while not q.empty():
a__ : List[Any] = []
while not q.empty():
a__ : Tuple = q.get()
print(node_dequeued.data , end="," )
if node_dequeued.left:
list_.append(node_dequeued.left )
if node_dequeued.right:
list_.append(node_dequeued.right )
print()
for node in list_:
q.put(__a )
def UpperCamelCase_ ( __a ) -> None:
if not isinstance(__a , __a ) or not node:
return
a__ : list[TreeNode] = []
a__ : Union[str, Any] = node
while n or stack:
while n: # start from root node, find its left child
print(n.data , end="," )
stack.append(__a )
a__ : List[Any] = n.left
# end of while means current node doesn't have left child
a__ : Optional[int] = stack.pop()
# start to traverse its right child
a__ : List[Any] = n.right
def UpperCamelCase_ ( __a ) -> None:
if not isinstance(__a , __a ) or not node:
return
a__ : list[TreeNode] = []
a__ : List[str] = node
while n or stack:
while n:
stack.append(__a )
a__ : List[str] = n.left
a__ : Dict = stack.pop()
print(n.data , end="," )
a__ : str = n.right
def UpperCamelCase_ ( __a ) -> None:
if not isinstance(__a , __a ) or not node:
return
a__, a__ : Union[str, Any] = [], []
a__ : Tuple = node
stacka.append(__a )
while stacka: # to find the reversed order of post order, store it in stack2
a__ : Union[str, Any] = stacka.pop()
if n.left:
stacka.append(n.left )
if n.right:
stacka.append(n.right )
stacka.append(__a )
while stacka: # pop up from stack2 will be the post order
print(stacka.pop().data , end="," )
def UpperCamelCase_ ( __a = "" , __a=50 , __a="*" ) -> str:
if not s:
return "\n" + width * char
a__, a__ : Any = divmod(width - len(__a ) - 2 , 2 )
return f'''{left * char} {s} {(left + extra) * char}'''
if __name__ == "__main__":
import doctest
doctest.testmod()
print(prompt("""Binary Tree Traversals"""))
UpperCamelCase : TreeNode = build_tree()
print(prompt("""Pre Order Traversal"""))
pre_order(node)
print(prompt() + """\n""")
print(prompt("""In Order Traversal"""))
in_order(node)
print(prompt() + """\n""")
print(prompt("""Post Order Traversal"""))
post_order(node)
print(prompt() + """\n""")
print(prompt("""Level Order Traversal"""))
level_order(node)
print(prompt() + """\n""")
print(prompt("""Actual Level Order Traversal"""))
level_order_actual(node)
print("""*""" * 50 + """\n""")
print(prompt("""Pre Order Traversal - Iteration Version"""))
pre_order_iter(node)
print(prompt() + """\n""")
print(prompt("""In Order Traversal - Iteration Version"""))
in_order_iter(node)
print(prompt() + """\n""")
print(prompt("""Post Order Traversal - Iteration Version"""))
post_order_iter(node)
print(prompt())
| 37 |
import os
from typing import Optional
import fsspec
from fsspec.archive import AbstractArchiveFileSystem
from fsspec.utils import DEFAULT_BLOCK_SIZE
class A__ ( A__ ):
"""simple docstring"""
_lowercase = ''
_lowercase = (
None # protocol passed in prefix to the url. ex: "gzip", for gzip://file.txt::http://foo.bar/file.txt.gz
)
_lowercase = None # compression type in fsspec. ex: "gzip"
_lowercase = None # extension of the filename to strip. ex: "".gz" to get file.txt from file.txt.gz
def __init__( self : List[str] , lowerCamelCase__ : str = "" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , **lowerCamelCase__ : List[str] ):
super().__init__(self , **lowerCamelCase__ )
# always open as "rb" since fsspec can then use the TextIOWrapper to make it work for "r" mode
a__ : str = fsspec.open(
lowerCamelCase__ , mode="rb" , protocol=lowerCamelCase__ , compression=self.compression , client_kwargs={
"requote_redirect_url": False, # see https://github.com/huggingface/datasets/pull/5459
"trust_env": True, # Enable reading proxy env variables.
**(target_options or {}).pop("client_kwargs" , {} ), # To avoid issues if it was already passed.
} , **(target_options or {}) , )
a__ : Optional[int] = os.path.basename(self.file.path.split("::" )[0] )
a__ : int = (
self.compressed_name[: self.compressed_name.rindex("." )]
if "." in self.compressed_name
else self.compressed_name
)
a__ : List[Any] = None
@classmethod
def _UpperCamelCase( cls : int , lowerCamelCase__ : int ):
# compressed file paths are always relative to the archive root
return super()._strip_protocol(lowerCamelCase__ ).lstrip("/" )
def _UpperCamelCase( self : Dict ):
if self.dir_cache is None:
a__ : Dict = {**self.file.fs.info(self.file.path ), "name": self.uncompressed_name}
a__ : int = {f["name"]: f}
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str ):
return self.file.open().read()
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : int=None , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[str]=None , **lowerCamelCase__ : Optional[Any] , ):
a__ : Optional[int] = self._strip_protocol(lowerCamelCase__ )
if mode != "rb":
raise ValueError(f'''Tried to read with mode {mode} on file {self.file.path} opened with mode \'rb\'''' )
return self.file.open()
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'bz2'
_lowercase = 'bz2'
_lowercase = '.bz2'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gzip'
_lowercase = 'gzip'
_lowercase = '.gz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'lz4'
_lowercase = 'lz4'
_lowercase = '.lz4'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'xz'
_lowercase = 'xz'
_lowercase = '.xz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'zstd'
_lowercase = 'zstd'
_lowercase = '.zst'
def __init__( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , lowerCamelCase__ : int = DEFAULT_BLOCK_SIZE , **lowerCamelCase__ : Tuple , ):
super().__init__(
fo=lowerCamelCase__ , mode=lowerCamelCase__ , target_protocol=lowerCamelCase__ , target_options=lowerCamelCase__ , block_size=lowerCamelCase__ , **lowerCamelCase__ , )
# We need to wrap the zstd decompressor to avoid this error in fsspec==2021.7.0 and zstandard==0.15.2:
#
# File "/Users/user/.virtualenvs/hf-datasets/lib/python3.7/site-packages/fsspec/core.py", line 145, in open
# out.close = close
# AttributeError: 'zstd.ZstdDecompressionReader' object attribute 'close' is read-only
#
# see https://github.com/intake/filesystem_spec/issues/725
a__ : Any = self.file.__enter__
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : str ):
a__ : List[Any] = file_
def __enter__( self : str ):
self._file.__enter__()
return self
def __exit__( self : int , *lowerCamelCase__ : List[str] , **lowerCamelCase__ : int ):
self._file.__exit__(*lowerCamelCase__ , **lowerCamelCase__ )
def __iter__( self : List[str] ):
return iter(self._file )
def _UpperCamelCase( self : Any ):
return next(self._file )
def __getattr__( self : Optional[Any] , lowerCamelCase__ : Tuple ):
return getattr(self._file , lowerCamelCase__ )
def fixed_enter(*lowerCamelCase__ : List[str] , **lowerCamelCase__ : str ):
return WrappedFile(_enter(*lowerCamelCase__ , **lowerCamelCase__ ) )
a__ : Any = fixed_enter
| 37 | 1 |
from ...configuration_utils import PretrainedConfig
from ...utils import logging
UpperCamelCase : Optional[Any] = logging.get_logger(__name__)
UpperCamelCase : str = {
"""EleutherAI/gpt-neox-20b""": """https://huggingface.co/EleutherAI/gpt-neox-20b/resolve/main/config.json""",
# See all GPTNeoX models at https://huggingface.co/models?filter=gpt_neox
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gpt_neox'
def __init__( self : int , lowerCamelCase__ : int=50_432 , lowerCamelCase__ : Optional[int]=6_144 , lowerCamelCase__ : Optional[Any]=44 , lowerCamelCase__ : Tuple=64 , lowerCamelCase__ : Union[str, Any]=24_576 , lowerCamelCase__ : List[str]="gelu" , lowerCamelCase__ : Union[str, Any]=0.25 , lowerCamelCase__ : Optional[int]=10_000 , lowerCamelCase__ : Union[str, Any]=0.0 , lowerCamelCase__ : List[Any]=0.0 , lowerCamelCase__ : int=0.1 , lowerCamelCase__ : Tuple=2_048 , lowerCamelCase__ : str=0.02 , lowerCamelCase__ : Any=1E-5 , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : Optional[Any]=0 , lowerCamelCase__ : Tuple=2 , lowerCamelCase__ : Optional[Any]=False , lowerCamelCase__ : str=True , lowerCamelCase__ : Optional[int]=None , **lowerCamelCase__ : Tuple , ):
super().__init__(bos_token_id=lowerCamelCase__ , eos_token_id=lowerCamelCase__ , **lowerCamelCase__ )
a__ : int = vocab_size
a__ : str = max_position_embeddings
a__ : Optional[int] = hidden_size
a__ : List[str] = num_hidden_layers
a__ : Tuple = num_attention_heads
a__ : int = intermediate_size
a__ : Optional[int] = hidden_act
a__ : str = rotary_pct
a__ : Optional[Any] = rotary_emb_base
a__ : Any = attention_dropout
a__ : Dict = hidden_dropout
a__ : Optional[int] = classifier_dropout
a__ : Dict = initializer_range
a__ : int = layer_norm_eps
a__ : Tuple = use_cache
a__ : Optional[Any] = tie_word_embeddings
a__ : List[str] = use_parallel_residual
a__ : str = rope_scaling
self._rope_scaling_validation()
if self.hidden_size % self.num_attention_heads != 0:
raise ValueError(
"The hidden size is not divisble by the number of attention heads! Make sure to update them!" )
def _UpperCamelCase( self : Optional[Any] ):
if self.rope_scaling is None:
return
if not isinstance(self.rope_scaling , lowerCamelCase__ ) or len(self.rope_scaling ) != 2:
raise ValueError(
"`rope_scaling` must be a dictionary with with two fields, `name` and `factor`, "
f'''got {self.rope_scaling}''' )
a__ : Dict = self.rope_scaling.get("type" , lowerCamelCase__ )
a__ : List[str] = self.rope_scaling.get("factor" , lowerCamelCase__ )
if rope_scaling_type is None or rope_scaling_type not in ["linear", "dynamic"]:
raise ValueError(
f'''`rope_scaling`\'s name field must be one of [\'linear\', \'dynamic\'], got {rope_scaling_type}''' )
if rope_scaling_factor is None or not isinstance(lowerCamelCase__ , lowerCamelCase__ ) or rope_scaling_factor <= 1.0:
raise ValueError(f'''`rope_scaling`\'s factor field must be an float > 1, got {rope_scaling_factor}''' )
| 37 |
import argparse
import os
import numpy as np
import tensorflow as tf
import torch
from transformers import BertModel
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Union[str, Any] = ("dense.weight", "attention.self.query", "attention.self.key", "attention.self.value")
a__ : Union[str, Any] = (
("layer.", "layer_"),
("word_embeddings.weight", "word_embeddings"),
("position_embeddings.weight", "position_embeddings"),
("token_type_embeddings.weight", "token_type_embeddings"),
(".", "/"),
("LayerNorm/weight", "LayerNorm/gamma"),
("LayerNorm/bias", "LayerNorm/beta"),
("weight", "kernel"),
)
if not os.path.isdir(__a ):
os.makedirs(__a )
a__ : Any = model.state_dict()
def to_tf_var_name(__a ):
for patt, repl in iter(__a ):
a__ : Tuple = name.replace(__a , __a )
return f'''bert/{name}'''
def create_tf_var(__a , __a , __a ):
a__ : Tuple = tf.dtypes.as_dtype(tensor.dtype )
a__ : Dict = tf.get_variable(dtype=__a , shape=tensor.shape , name=__a , initializer=tf.zeros_initializer() )
session.run(tf.variables_initializer([tf_var] ) )
session.run(__a )
return tf_var
tf.reset_default_graph()
with tf.Session() as session:
for var_name in state_dict:
a__ : int = to_tf_var_name(__a )
a__ : Union[str, Any] = state_dict[var_name].numpy()
if any(x in var_name for x in tensors_to_transpose ):
a__ : int = torch_tensor.T
a__ : Optional[Any] = create_tf_var(tensor=__a , name=__a , session=__a )
tf.keras.backend.set_value(__a , __a )
a__ : int = session.run(__a )
print(f'''Successfully created {tf_name}: {np.allclose(__a , __a )}''' )
a__ : Any = tf.train.Saver(tf.trainable_variables() )
saver.save(__a , os.path.join(__a , model_name.replace("-" , "_" ) + ".ckpt" ) )
def UpperCamelCase_ ( __a=None ) -> int:
a__ : Dict = argparse.ArgumentParser()
parser.add_argument("--model_name" , type=__a , required=__a , help="model name e.g. bert-base-uncased" )
parser.add_argument(
"--cache_dir" , type=__a , default=__a , required=__a , help="Directory containing pytorch model" )
parser.add_argument("--pytorch_model_path" , type=__a , required=__a , help="/path/to/<pytorch-model-name>.bin" )
parser.add_argument("--tf_cache_dir" , type=__a , required=__a , help="Directory in which to save tensorflow model" )
a__ : Optional[Any] = parser.parse_args(__a )
a__ : Tuple = BertModel.from_pretrained(
pretrained_model_name_or_path=args.model_name , state_dict=torch.load(args.pytorch_model_path ) , cache_dir=args.cache_dir , )
convert_pytorch_checkpoint_to_tf(model=__a , ckpt_dir=args.tf_cache_dir , model_name=args.model_name )
if __name__ == "__main__":
main()
| 37 | 1 |
UpperCamelCase : str = """
# Installazione di Transformers
! pip install transformers datasets
# Per installare dalla fonte invece dell'ultima versione rilasciata, commenta il comando sopra e
# rimuovi la modalità commento al comando seguente.
# ! pip install git+https://github.com/huggingface/transformers.git
"""
UpperCamelCase : Tuple = [{"""type""": """code""", """content""": INSTALL_CONTENT}]
UpperCamelCase : Optional[int] = {
"""{processor_class}""": """FakeProcessorClass""",
"""{model_class}""": """FakeModelClass""",
"""{object_class}""": """FakeObjectClass""",
}
| 37 |
import inspect
import unittest
from huggingface_hub import hf_hub_download
from transformers import ASTConfig
from transformers.testing_utils import require_torch, require_torchaudio, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_torchaudio_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import ASTForAudioClassification, ASTModel
from transformers.models.audio_spectrogram_transformer.modeling_audio_spectrogram_transformer import (
AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
)
if is_torchaudio_available():
import torchaudio
from transformers import ASTFeatureExtractor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[str] , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[Any]=2 , lowerCamelCase__ : Any=24 , lowerCamelCase__ : Optional[Any]=16 , lowerCamelCase__ : int=True , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[Any]=32 , lowerCamelCase__ : List[str]=5 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Optional[Any]=37 , lowerCamelCase__ : Any="gelu" , lowerCamelCase__ : Union[str, Any]=0.1 , lowerCamelCase__ : Optional[int]=0.1 , lowerCamelCase__ : str=10 , lowerCamelCase__ : Optional[Any]=0.02 , lowerCamelCase__ : str=None , lowerCamelCase__ : List[str]=2 , lowerCamelCase__ : Optional[Any]=2 , ):
a__ : str = parent
a__ : Any = batch_size
a__ : Dict = patch_size
a__ : List[Any] = max_length
a__ : str = num_mel_bins
a__ : Optional[Any] = is_training
a__ : Optional[int] = use_labels
a__ : List[Any] = hidden_size
a__ : str = num_hidden_layers
a__ : Any = num_attention_heads
a__ : Union[str, Any] = intermediate_size
a__ : List[str] = hidden_act
a__ : str = hidden_dropout_prob
a__ : Tuple = attention_probs_dropout_prob
a__ : List[Any] = type_sequence_label_size
a__ : Any = initializer_range
a__ : str = scope
a__ : List[str] = frequency_stride
a__ : Union[str, Any] = time_stride
# in AST, the seq length equals the number of patches + 2 (we add 2 for the [CLS] and distillation tokens)
a__ : List[Any] = (self.num_mel_bins - self.patch_size) // self.frequency_stride + 1
a__ : List[str] = (self.max_length - self.patch_size) // self.time_stride + 1
a__ : Tuple = frequency_out_dimension * time_out_dimension
a__ : List[str] = num_patches + 2
def _UpperCamelCase( self : List[str] ):
a__ : Any = floats_tensor([self.batch_size, self.max_length, self.num_mel_bins] )
a__ : List[Any] = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : List[str] = self.get_config()
return config, input_values, labels
def _UpperCamelCase( self : Optional[int] ):
return ASTConfig(
patch_size=self.patch_size , max_length=self.max_length , num_mel_bins=self.num_mel_bins , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , frequency_stride=self.frequency_stride , time_stride=self.time_stride , )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : int , lowerCamelCase__ : Optional[int] ):
a__ : List[Any] = ASTModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Dict = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : str ):
a__ : Dict = self.prepare_config_and_inputs()
(
(
a__
), (
a__
), (
a__
),
) : Optional[int] = config_and_inputs
a__ : List[Any] = {"input_values": input_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(
ASTModel,
ASTForAudioClassification,
)
if is_torch_available()
else ()
)
_lowercase = (
{'audio-classification': ASTForAudioClassification, 'feature-extraction': ASTModel}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
if pipeline_test_casse_name == "AudioClassificationPipelineTests":
return True
return False
def _UpperCamelCase( self : str ):
a__ : str = ASTModelTester(self )
a__ : Any = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[str] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="AST does not use inputs_embeds" )
def _UpperCamelCase( self : List[str] ):
pass
def _UpperCamelCase( self : Optional[int] ):
a__, a__ : str = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Any = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Union[str, Any] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : Tuple ):
a__, a__ : List[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Dict = model_class(lowerCamelCase__ )
a__ : Optional[int] = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Optional[Any] = ["input_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
for model_name in AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Union[str, Any] = ASTModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : Optional[int] = hf_hub_download(
repo_id="nielsr/audio-spectogram-transformer-checkpoint" , filename="sample_audio.flac" , repo_type="dataset" )
a__, a__ : List[str] = torchaudio.load(__a )
return audio, sampling_rate
@require_torch
@require_torchaudio
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : List[str] ):
return (
ASTFeatureExtractor.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" )
if is_torchaudio_available()
else None
)
@slow
def _UpperCamelCase( self : Optional[int] ):
a__ : int = self.default_feature_extractor
a__ : Optional[Any] = ASTForAudioClassification.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" ).to(lowerCamelCase__ )
a__ : Any = self.default_feature_extractor
a__, a__ : Dict = prepare_audio()
a__ : str = audio.squeeze().numpy()
a__ : Any = feature_extractor(lowerCamelCase__ , sampling_rate=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(**lowerCamelCase__ )
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 527) )
self.assertEqual(outputs.logits.shape , lowerCamelCase__ )
a__ : List[str] = torch.tensor([-0.8760, -7.0042, -8.6602] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(outputs.logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
| 37 | 1 |
def UpperCamelCase_ ( __a , __a ) -> Tuple:
a__ : Optional[int] = [0 for i in range(r + 1 )]
# nc0 = 1
a__ : Union[str, Any] = 1
for i in range(1 , n + 1 ):
# to compute current row from previous row.
a__ : Any = min(__a , __a )
while j > 0:
c[j] += c[j - 1]
j -= 1
return c[r]
print(binomial_coefficient(n=10, r=5))
| 37 |
import pickle
import shutil
import tempfile
import unittest
from transformers import SPIECE_UNDERLINE, XGLMTokenizer, XGLMTokenizerFast
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers, slow
from transformers.utils import cached_property
from ...test_tokenization_common import TokenizerTesterMixin
UpperCamelCase : Optional[Any] = get_tests_dir("""fixtures/test_sentencepiece.model""")
@require_sentencepiece
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = XGLMTokenizer
_lowercase = XGLMTokenizerFast
_lowercase = True
_lowercase = True
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# We have a SentencePiece fixture for testing
a__ : str = XGLMTokenizer(lowerCamelCase__ , keep_accents=lowerCamelCase__ )
tokenizer.save_pretrained(self.tmpdirname )
def _UpperCamelCase( self : List[Any] ):
a__ : int = "<pad>"
a__ : Union[str, Any] = 1
self.assertEqual(self.get_tokenizer()._convert_token_to_id(lowerCamelCase__ ) , lowerCamelCase__ )
self.assertEqual(self.get_tokenizer()._convert_id_to_token(lowerCamelCase__ ) , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : List[str] = list(self.get_tokenizer().get_vocab().keys() )
self.assertEqual(vocab_keys[0] , "<s>" )
self.assertEqual(vocab_keys[1] , "<pad>" )
self.assertEqual(len(lowerCamelCase__ ) , 1_008 )
def _UpperCamelCase( self : Dict ):
self.assertEqual(self.get_tokenizer().vocab_size , 1_008 )
def _UpperCamelCase( self : Optional[int] ):
a__ : str = XGLMTokenizer(lowerCamelCase__ , keep_accents=lowerCamelCase__ )
a__ : List[str] = tokenizer.tokenize("This is a test" )
self.assertListEqual(lowerCamelCase__ , ["▁This", "▁is", "▁a", "▁t", "est"] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
a__ : Any = tokenizer.tokenize("I was born in 92000, and this is falsé." )
self.assertListEqual(
lowerCamelCase__ , [
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"9",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"é",
".",
] , )
a__ : List[str] = tokenizer.convert_tokens_to_ids(lowerCamelCase__ )
self.assertListEqual(
lowerCamelCase__ , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
a__ : Dict = tokenizer.convert_ids_to_tokens(lowerCamelCase__ )
self.assertListEqual(
lowerCamelCase__ , [
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"<unk>",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"<unk>",
".",
] , )
@cached_property
def _UpperCamelCase( self : Dict ):
return XGLMTokenizer.from_pretrained("facebook/xglm-564M" )
def _UpperCamelCase( self : Union[str, Any] ):
with tempfile.NamedTemporaryFile() as f:
shutil.copyfile(lowerCamelCase__ , f.name )
a__ : Any = XGLMTokenizer(f.name , keep_accents=lowerCamelCase__ )
a__ : List[str] = pickle.dumps(lowerCamelCase__ )
pickle.loads(lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
if not self.test_rust_tokenizer:
return
a__ : Any = self.get_tokenizer()
a__ : Optional[Any] = self.get_rust_tokenizer()
a__ : Tuple = "I was born in 92000, and this is falsé."
a__ : List[str] = tokenizer.tokenize(lowerCamelCase__ )
a__ : Union[str, Any] = rust_tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : Optional[int] = tokenizer.encode(lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
a__ : Union[str, Any] = rust_tokenizer.encode(lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : List[str] = self.get_rust_tokenizer()
a__ : Tuple = tokenizer.encode(lowerCamelCase__ )
a__ : Optional[Any] = rust_tokenizer.encode(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = "Hello World!"
a__ : List[str] = [2, 31_227, 4_447, 35]
self.assertListEqual(lowerCamelCase__ , self.big_tokenizer.encode(lowerCamelCase__ ) )
@slow
def _UpperCamelCase( self : Union[str, Any] ):
a__ : Optional[int] = (
"This is a very long text with a lot of weird characters, such as: . , ~ ? ( ) \" [ ] ! : - . Also we will"
" add words that should not exsist and be tokenized to unk, such as saoneuhaoesuth"
)
# fmt: off
a__ : Union[str, Any] = [2, 1_018, 67, 11, 1_988, 2_617, 5_631, 278, 11, 3_407, 48, 71_630, 28_085, 4, 3_234, 157, 13, 6, 5, 6, 4, 3_526, 768, 15, 659, 57, 298, 3_983, 864, 129, 21, 6, 5, 13_675, 377, 652, 7_580, 10_341, 155, 2_817, 422, 1_666, 7, 1_674, 53, 113, 202_277, 17_892, 33, 60, 87, 4, 3_234, 157, 61, 2_667, 52_376, 19, 88, 23, 735]
# fmt: on
self.assertListEqual(lowerCamelCase__ , self.big_tokenizer.encode(lowerCamelCase__ ) )
@slow
def _UpperCamelCase( self : List[Any] ):
# fmt: off
a__ : Optional[int] = {
"input_ids": [[2, 108_825, 1_163, 15, 88_010, 473, 15_898, 157, 13_672, 1_857, 312, 8, 238_021, 1_163, 53, 13_672, 1_857, 312, 8, 53_283, 182_396, 8, 18_566, 16, 36_733, 4_101, 8, 230, 244_017, 122_553, 7, 15, 132_597, 4, 293, 12_511, 7_610, 4, 3_414, 132_597, 9, 4, 32_361, 362, 4, 734, 28_512, 32_569, 18, 4, 32_361, 26_096, 14_982, 73, 18_715, 21_433, 235_261, 15, 492, 12_427, 16, 53, 18_715, 21_433, 65_454, 15, 23_659, 563, 16, 278, 597, 2_843, 595, 7_931, 182_396, 64_186, 22, 886, 595, 132_981, 53, 25_540, 3_449, 43_982, 39_901, 5_951, 878, 330, 4, 27_694, 80_269, 312, 53, 6_517, 11_780, 611, 20_408, 5], [2, 6, 132_597, 67, 42_897, 33, 592, 8, 163_729, 25_540, 361, 136_997, 109_514, 173_230, 7, 501, 60, 102_913, 196, 5_631, 235, 63_243, 473, 6, 231_757, 74, 5_277, 7_905, 53, 3_095, 37_317, 22, 454, 183_874, 5], [2, 268, 31_298, 46_530, 6, 132_935, 43_831, 7, 597, 32, 24, 3_688, 9_865, 5]],
"attention_mask": [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]
} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=lowerCamelCase__ , model_name="facebook/xglm-564M" , padding=lowerCamelCase__ , )
| 37 | 1 |
from statistics import mean, stdev
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : List[str] = min(__a )
a__ : str = max(__a )
# normalize data
return [round((x - x_min) / (x_max - x_min) , __a ) for x in data]
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : str = mean(__a )
a__ : List[str] = stdev(__a )
# standardize data
return [round((x - mu) / (sigma) , __a ) for x in data]
| 37 |
import argparse
import requests
import torch
# pip3 install salesforce-lavis
# I'm actually installing a slightly modified version: pip3 install git+https://github.com/nielsrogge/LAVIS.git@fix_lavis_float32 (there's also the fix_lavis branch)
# also note: to convert Vicuna checkpoints, we had to include /home/niels/python_projects/checkpoints/FastChat/vicuna-7b in lavis/configs/models/blip2/blip2_instruct_vicuna7b.yaml
# same for Vicuna-13b
from lavis.models import load_model_and_preprocess
from PIL import Image
from transformers import (
AutoTokenizer,
BlipImageProcessor,
InstructBlipConfig,
InstructBlipForConditionalGeneration,
InstructBlipProcessor,
InstructBlipQFormerConfig,
InstructBlipVisionConfig,
LlamaConfig,
LlamaTokenizerFast,
TaConfig,
TaTokenizerFast,
)
from transformers.utils.constants import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD
def UpperCamelCase_ ( ) -> int:
a__ : int = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
a__ : Optional[Any] = Image.open(requests.get(__a , stream=__a ).raw ).convert("RGB" )
return image
def UpperCamelCase_ ( __a ) -> Optional[Any]:
a__ : Any = []
# fmt: off
# vision encoder
rename_keys.append(("visual_encoder.cls_token", "vision_model.embeddings.class_embedding") )
rename_keys.append(("visual_encoder.pos_embed", "vision_model.embeddings.position_embedding") )
rename_keys.append(("visual_encoder.patch_embed.proj.weight", "vision_model.embeddings.patch_embedding.weight") )
rename_keys.append(("visual_encoder.patch_embed.proj.bias", "vision_model.embeddings.patch_embedding.bias") )
rename_keys.append(("ln_vision.weight", "vision_model.post_layernorm.weight") )
rename_keys.append(("ln_vision.bias", "vision_model.post_layernorm.bias") )
for i in range(config.vision_config.num_hidden_layers ):
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.weight''', f'''vision_model.encoder.layers.{i}.layer_norm1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.bias''', f'''vision_model.encoder.layers.{i}.layer_norm1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.weight''', f'''vision_model.encoder.layers.{i}.layer_norm2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.bias''', f'''vision_model.encoder.layers.{i}.layer_norm2.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.qkv.weight''', f'''vision_model.encoder.layers.{i}.self_attn.qkv.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.weight''', f'''vision_model.encoder.layers.{i}.self_attn.projection.weight''',) )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.bias''', f'''vision_model.encoder.layers.{i}.self_attn.projection.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc2.bias''') )
# QFormer
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.weight", "qformer.embeddings.layernorm.weight") )
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.bias", "qformer.embeddings.layernorm.bias") )
# fmt: on
return rename_keys
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Union[str, Any] = dct.pop(__a )
a__ : List[str] = val
def UpperCamelCase_ ( __a , __a ) -> Optional[Any]:
for i in range(config.vision_config.num_hidden_layers ):
# read in original q and v biases
a__ : Any = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.q_bias''' )
a__ : Tuple = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.v_bias''' )
# next, set bias in the state dict
a__ : str = torch.cat((q_bias, torch.zeros_like(__a , requires_grad=__a ), v_bias) )
a__ : int = qkv_bias
def UpperCamelCase_ ( __a ) -> Dict:
a__ : Tuple = 364 if "coco" in model_name else 224
a__ : int = InstructBlipVisionConfig(image_size=__a ).to_dict()
# make sure the models have proper bos_token_id and eos_token_id set (important for generation)
# seems like flan-T5 models don't have bos_token_id properly set?
if "t5-xl" in model_name:
a__ : Tuple = TaConfig.from_pretrained("google/flan-t5-xl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "t5-xxl" in model_name:
a__ : Dict = TaConfig.from_pretrained("google/flan-t5-xxl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "vicuna-7b" in model_name:
a__ : List[Any] = LlamaConfig.from_pretrained("decapoda-research/llama-7b-hf" , vocab_size=32_001 ).to_dict()
elif "vicuna-13b" in model_name:
a__ : Optional[int] = LlamaConfig.from_pretrained("decapoda-research/llama-13b-hf" , vocab_size=32_001 ).to_dict()
else:
raise ValueError("Model name not supported" )
# the authors add one special "[DEC]" token to the vocab of Q-Former, hence vocab size = 30522 + 1
a__ : Optional[Any] = InstructBlipQFormerConfig(vocab_size=30_523 ).to_dict()
a__ : Any = InstructBlipConfig(vision_config=__a , text_config=__a , qformer_config=__a )
return config, image_size
@torch.no_grad()
def UpperCamelCase_ ( __a , __a=None , __a=False ) -> int:
a__ : Tuple = AutoTokenizer.from_pretrained("bert-base-uncased" , truncation_side="left" )
qformer_tokenizer.add_special_tokens({"bos_token": "[DEC]"} )
if "t5" in model_name:
a__ : List[Any] = TaTokenizerFast.from_pretrained("google/flan-t5-xl" , truncation_side="left" )
elif "vicuna" in model_name:
# the following was used in the original implementation:
# tokenizer = LlamaTokenizer.from_pretrained("huggyllama/llama-7b", use_fast=False, truncation_side="left")
# tokenizer.add_special_tokens({"pad_token": "[PAD]"})
# tokenizer.add_special_tokens({"bos_token": "</s>"})
# tokenizer.add_special_tokens({"eos_token": "</s>"})
# tokenizer.add_special_tokens({"unk_token": "</s>"})
a__ : Union[str, Any] = LlamaTokenizerFast.from_pretrained(
"huggyllama/llama-7b" , truncation_side="left" , bos_token="</s>" , unk_token="</s>" )
tokenizer.add_special_tokens({"pad_token": "[PAD]"} )
a__, a__ : List[str] = get_blipa_config(__a )
a__ : Any = InstructBlipForConditionalGeneration(__a ).eval()
a__ : Dict = {
"instructblip-vicuna-7b": ("blip2_vicuna_instruct", "vicuna7b"),
"instructblip-vicuna-13b": ("blip2_vicuna_instruct", "vicuna13b"),
"instructblip-flan-t5-xl": ("blip2_t5_instruct", "flant5xl"),
"instructblip-flan-t5-xxl": ("blip2_t5_instruct", "flant5xxl"),
}
a__, a__ : Dict = model_name_to_original[model_name]
# load original model
print("Loading original model..." )
a__ : Optional[Any] = "cuda:1" if torch.cuda.is_available() else "cpu"
a__ : List[Any] = "cuda:2" if torch.cuda.is_available() else "cpu"
a__, a__, a__ : Tuple = load_model_and_preprocess(
name=__a , model_type=__a , is_eval=__a , device=__a )
original_model.eval()
print("Done!" )
# update state dict keys
a__ : Dict = original_model.state_dict()
a__ : Optional[int] = create_rename_keys(__a )
for src, dest in rename_keys:
rename_key(__a , __a , __a )
# some keys can be renamed efficiently
for key, val in state_dict.copy().items():
a__ : Optional[int] = state_dict.pop(__a )
if key.startswith("Qformer.bert" ):
a__ : List[Any] = key.replace("Qformer.bert" , "qformer" )
if "attention.self" in key:
a__ : Any = key.replace("self" , "attention" )
if "llm_proj" in key:
a__ : Dict = key.replace("llm_proj" , "language_projection" )
if "t5_proj" in key:
a__ : int = key.replace("t5_proj" , "language_projection" )
if key.startswith("llm_model" ):
a__ : List[str] = key.replace("llm_model" , "language_model" )
if key.startswith("t5" ):
a__ : str = key.replace("t5" , "language" )
a__ : Dict = val
# read in qv biases
read_in_q_v_bias(__a , __a )
# note: weights get loaded in torch.float32 by default
hf_model.load_state_dict(__a , strict=__a )
a__ : Union[str, Any] = load_demo_image()
a__ : int = "What is unusual about this image?"
# create processor
a__ : Any = BlipImageProcessor(
size={"height": image_size, "width": image_size} , image_mean=__a , image_std=__a )
a__ : Tuple = InstructBlipProcessor(
image_processor=__a , tokenizer=__a , qformer_tokenizer=__a , )
a__ : Tuple = processor(images=__a , text=__a , return_tensors="pt" ).to(__a )
# make sure processor creates exact same pixel values
a__ : Optional[int] = vis_processors["eval"](__a ).unsqueeze(0 ).to(__a )
a__ : Optional[Any] = inputs.pixel_values
assert torch.allclose(original_pixel_values.to(pixel_values.device ) , __a )
original_model.to(__a )
hf_model.to(__a )
with torch.no_grad():
if "vicuna" in model_name:
a__ : str = original_model({"image": original_pixel_values, "text_input": [prompt]} ).logits
a__ : List[str] = hf_model(**__a ).logits
else:
a__ : List[Any] = original_model(
{"image": original_pixel_values, "text_input": [prompt], "text_output": ["\n"]} ).logits
a__ : str = tokenizer("\n" , return_tensors="pt" ).input_ids.to(__a )
a__ : Dict = label_input_ids.masked_fill(label_input_ids == tokenizer.pad_token_id , -100 )
a__ : Any = hf_model(**__a , labels=__a ).logits
print("First values of original logits:" , original_logits[0, :3, :3] )
print("First values of HF logits:" , logits[0, :3, :3] )
# assert values
assert original_logits.shape == logits.shape
a__ : Tuple = 1e-4 if "vicuna" in model_name else 1e-5
assert torch.allclose(original_logits.to(logits.device ) , __a , atol=__a )
print("Looks ok!" )
print("Generating with original model..." )
a__ : Tuple = original_model.generate({"image": original_pixel_values, "prompt": prompt} , num_beams=5 )
# important: we need to cast the weights of the HF model to the appropriate type
print("Generating with HF model..." )
a__ : int = hf_model.generate(
**__a , do_sample=__a , num_beams=5 , max_length=256 , min_length=1 , top_p=0.9 , repetition_penalty=1.5 , length_penalty=1.0 , temperature=1 , )
if "vicuna" in model_name:
# convert output id 0 to 2 (eos_token_id)
# TODO add this in the generate method?
a__ : int = 2
print("Original generation:" , __a )
a__ : str = processor.batch_decode(__a , skip_special_tokens=__a )
a__ : str = [text.strip() for text in output_text]
print("HF generation:" , __a )
if pytorch_dump_folder_path is not None:
processor.save_pretrained(__a )
hf_model.save_pretrained(__a )
if push_to_hub:
processor.push_to_hub(f'''Salesforce/{model_name}''' )
hf_model.push_to_hub(f'''Salesforce/{model_name}''' )
if __name__ == "__main__":
UpperCamelCase : Any = argparse.ArgumentParser()
UpperCamelCase : Optional[int] = [
"""instructblip-vicuna-7b""",
"""instructblip-vicuna-13b""",
"""instructblip-flan-t5-xl""",
"""instructblip-flan-t5-xxl""",
]
parser.add_argument(
"""--model_name""",
default="""instructblip-flan-t5-xl""",
choices=choices,
type=str,
help="""Path to hf config.json of model to convert""",
)
parser.add_argument("""--pytorch_dump_folder_path""", default=None, type=str, help="""Path to the output PyTorch model.""")
parser.add_argument(
"""--push_to_hub""",
action="""store_true""",
help="""Whether to push the model and processor to the hub after converting""",
)
UpperCamelCase : Dict = parser.parse_args()
convert_blipa_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 37 | 1 |
import argparse
import json
from pathlib import Path
import requests
import timm
import torch
from huggingface_hub import hf_hub_download
from PIL import Image
from transformers import DeiTConfig, DeiTForImageClassificationWithTeacher, DeiTImageProcessor
from transformers.utils import logging
logging.set_verbosity_info()
UpperCamelCase : str = logging.get_logger(__name__)
def UpperCamelCase_ ( __a , __a=False ) -> Dict:
a__ : Dict = []
for i in range(config.num_hidden_layers ):
# encoder layers: output projection, 2 feedforward neural networks and 2 layernorms
rename_keys.append((f'''blocks.{i}.norm1.weight''', f'''deit.encoder.layer.{i}.layernorm_before.weight''') )
rename_keys.append((f'''blocks.{i}.norm1.bias''', f'''deit.encoder.layer.{i}.layernorm_before.bias''') )
rename_keys.append((f'''blocks.{i}.attn.proj.weight''', f'''deit.encoder.layer.{i}.attention.output.dense.weight''') )
rename_keys.append((f'''blocks.{i}.attn.proj.bias''', f'''deit.encoder.layer.{i}.attention.output.dense.bias''') )
rename_keys.append((f'''blocks.{i}.norm2.weight''', f'''deit.encoder.layer.{i}.layernorm_after.weight''') )
rename_keys.append((f'''blocks.{i}.norm2.bias''', f'''deit.encoder.layer.{i}.layernorm_after.bias''') )
rename_keys.append((f'''blocks.{i}.mlp.fc1.weight''', f'''deit.encoder.layer.{i}.intermediate.dense.weight''') )
rename_keys.append((f'''blocks.{i}.mlp.fc1.bias''', f'''deit.encoder.layer.{i}.intermediate.dense.bias''') )
rename_keys.append((f'''blocks.{i}.mlp.fc2.weight''', f'''deit.encoder.layer.{i}.output.dense.weight''') )
rename_keys.append((f'''blocks.{i}.mlp.fc2.bias''', f'''deit.encoder.layer.{i}.output.dense.bias''') )
# projection layer + position embeddings
rename_keys.extend(
[
("cls_token", "deit.embeddings.cls_token"),
("dist_token", "deit.embeddings.distillation_token"),
("patch_embed.proj.weight", "deit.embeddings.patch_embeddings.projection.weight"),
("patch_embed.proj.bias", "deit.embeddings.patch_embeddings.projection.bias"),
("pos_embed", "deit.embeddings.position_embeddings"),
] )
if base_model:
# layernorm + pooler
rename_keys.extend(
[
("norm.weight", "layernorm.weight"),
("norm.bias", "layernorm.bias"),
("pre_logits.fc.weight", "pooler.dense.weight"),
("pre_logits.fc.bias", "pooler.dense.bias"),
] )
# if just the base model, we should remove "deit" from all keys that start with "deit"
a__ : List[str] = [(pair[0], pair[1][4:]) if pair[1].startswith("deit" ) else pair for pair in rename_keys]
else:
# layernorm + classification heads
rename_keys.extend(
[
("norm.weight", "deit.layernorm.weight"),
("norm.bias", "deit.layernorm.bias"),
("head.weight", "cls_classifier.weight"),
("head.bias", "cls_classifier.bias"),
("head_dist.weight", "distillation_classifier.weight"),
("head_dist.bias", "distillation_classifier.bias"),
] )
return rename_keys
def UpperCamelCase_ ( __a , __a , __a=False ) -> Optional[int]:
for i in range(config.num_hidden_layers ):
if base_model:
a__ : Optional[Any] = ""
else:
a__ : List[Any] = "deit."
# read in weights + bias of input projection layer (in timm, this is a single matrix + bias)
a__ : List[Any] = state_dict.pop(f'''blocks.{i}.attn.qkv.weight''' )
a__ : Union[str, Any] = state_dict.pop(f'''blocks.{i}.attn.qkv.bias''' )
# next, add query, keys and values (in that order) to the state dict
a__ : List[str] = in_proj_weight[
: config.hidden_size, :
]
a__ : List[Any] = in_proj_bias[: config.hidden_size]
a__ : Optional[int] = in_proj_weight[
config.hidden_size : config.hidden_size * 2, :
]
a__ : Optional[int] = in_proj_bias[
config.hidden_size : config.hidden_size * 2
]
a__ : Tuple = in_proj_weight[
-config.hidden_size :, :
]
a__ : Union[str, Any] = in_proj_bias[-config.hidden_size :]
def UpperCamelCase_ ( __a , __a , __a ) -> int:
a__ : List[str] = dct.pop(__a )
a__ : List[Any] = val
def UpperCamelCase_ ( ) -> Optional[Any]:
a__ : str = "http://images.cocodataset.org/val2017/000000039769.jpg"
a__ : Dict = Image.open(requests.get(__a , stream=__a ).raw )
return im
@torch.no_grad()
def UpperCamelCase_ ( __a , __a ) -> List[str]:
a__ : str = DeiTConfig()
# all deit models have fine-tuned heads
a__ : Tuple = False
# dataset (fine-tuned on ImageNet 2012), patch_size and image_size
a__ : Optional[Any] = 1_000
a__ : List[str] = "huggingface/label-files"
a__ : int = "imagenet-1k-id2label.json"
a__ : List[str] = json.load(open(hf_hub_download(__a , __a , repo_type="dataset" ) , "r" ) )
a__ : Union[str, Any] = {int(__a ): v for k, v in idalabel.items()}
a__ : Optional[Any] = idalabel
a__ : List[str] = {v: k for k, v in idalabel.items()}
a__ : Optional[Any] = int(deit_name[-6:-4] )
a__ : Tuple = int(deit_name[-3:] )
# size of the architecture
if deit_name[9:].startswith("tiny" ):
a__ : Dict = 192
a__ : Union[str, Any] = 768
a__ : List[str] = 12
a__ : Union[str, Any] = 3
elif deit_name[9:].startswith("small" ):
a__ : int = 384
a__ : int = 1_536
a__ : Optional[int] = 12
a__ : int = 6
if deit_name[9:].startswith("base" ):
pass
elif deit_name[4:].startswith("large" ):
a__ : List[str] = 1_024
a__ : List[str] = 4_096
a__ : int = 24
a__ : Any = 16
# load original model from timm
a__ : List[Any] = timm.create_model(__a , pretrained=__a )
timm_model.eval()
# load state_dict of original model, remove and rename some keys
a__ : Dict = timm_model.state_dict()
a__ : Optional[Any] = create_rename_keys(__a , __a )
for src, dest in rename_keys:
rename_key(__a , __a , __a )
read_in_q_k_v(__a , __a , __a )
# load HuggingFace model
a__ : Optional[int] = DeiTForImageClassificationWithTeacher(__a ).eval()
model.load_state_dict(__a )
# Check outputs on an image, prepared by DeiTImageProcessor
a__ : int = int(
(256 / 224) * config.image_size ) # to maintain same ratio w.r.t. 224 images, see https://github.com/facebookresearch/deit/blob/ab5715372db8c6cad5740714b2216d55aeae052e/datasets.py#L103
a__ : Tuple = DeiTImageProcessor(size=__a , crop_size=config.image_size )
a__ : Optional[int] = image_processor(images=prepare_img() , return_tensors="pt" )
a__ : Optional[int] = encoding["pixel_values"]
a__ : Any = model(__a )
a__ : List[str] = timm_model(__a )
assert timm_logits.shape == outputs.logits.shape
assert torch.allclose(__a , outputs.logits , atol=1e-3 )
Path(__a ).mkdir(exist_ok=__a )
print(f'''Saving model {deit_name} to {pytorch_dump_folder_path}''' )
model.save_pretrained(__a )
print(f'''Saving image processor to {pytorch_dump_folder_path}''' )
image_processor.save_pretrained(__a )
if __name__ == "__main__":
UpperCamelCase : Tuple = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"""--deit_name""",
default="""vit_deit_base_distilled_patch16_224""",
type=str,
help="""Name of the DeiT timm model you'd like to convert.""",
)
parser.add_argument(
"""--pytorch_dump_folder_path""", default=None, type=str, help="""Path to the output PyTorch model directory."""
)
UpperCamelCase : Optional[int] = parser.parse_args()
convert_deit_checkpoint(args.deit_name, args.pytorch_dump_folder_path)
| 37 |
def UpperCamelCase_ ( __a , __a ) -> Tuple:
a__ : Optional[int] = [0 for i in range(r + 1 )]
# nc0 = 1
a__ : Union[str, Any] = 1
for i in range(1 , n + 1 ):
# to compute current row from previous row.
a__ : Any = min(__a , __a )
while j > 0:
c[j] += c[j - 1]
j -= 1
return c[r]
print(binomial_coefficient(n=10, r=5))
| 37 | 1 |
def UpperCamelCase_ ( __a ) -> bool:
return sum(i for i in range(1 , number // 2 + 1 ) if number % i == 0 ) == number
if __name__ == "__main__":
print("""Program to check whether a number is a Perfect number or not...""")
UpperCamelCase : Union[str, Any] = int(input("""Enter number: """).strip())
print(f"""{number} is {"" if perfect(number) else "not "}a Perfect Number.""")
| 37 |
import json
from typing import Dict, List, Optional, Tuple, Union
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import PaddingStrategy, logging
from .tokenization_led import LEDTokenizer
UpperCamelCase : Union[str, Any] = logging.get_logger(__name__)
UpperCamelCase : Optional[Any] = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Optional[Any] = {
"""vocab_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json""",
},
"""merges_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt""",
},
"""tokenizer_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json""",
},
}
UpperCamelCase : Dict = {
"""allenai/led-base-16384""": 1_6384,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = LEDTokenizer
_lowercase = ['input_ids', 'attention_mask']
def __init__( self : Tuple , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[str]=None , lowerCamelCase__ : Any=None , lowerCamelCase__ : int="replace" , lowerCamelCase__ : Union[str, Any]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Tuple="</s>" , lowerCamelCase__ : Optional[int]="<s>" , lowerCamelCase__ : str="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Any="<mask>" , lowerCamelCase__ : Optional[int]=False , lowerCamelCase__ : int=True , **lowerCamelCase__ : Union[str, Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : List[str] = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : Optional[Any] = add_prefix_space
a__ : List[str] = pre_tok_class(**lowerCamelCase__ )
a__ : Optional[int] = add_prefix_space
# the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__`
a__ : Any = "post_processor"
a__ : str = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Any = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Optional[Any] = tuple(state["sep"] )
if "cls" in state:
a__ : Optional[Any] = tuple(state["cls"] )
a__ : Optional[int] = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Dict = add_prefix_space
a__ : int = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : List[Any] = trim_offsets
a__ : List[str] = True
if changes_to_apply:
a__ : int = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : int = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
# Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Union[str, Any] ):
a__ : Any = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : Union[str, Any] = value
def _UpperCamelCase( self : Any , *lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : List[str] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
if is_split_into_words and not self.add_prefix_space:
raise ValueError(
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs." )
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Any , *lowerCamelCase__ : Dict , **lowerCamelCase__ : Optional[Any] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
if is_split_into_words and not self.add_prefix_space:
raise ValueError(
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs." )
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : List[str] = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[Any]=None ):
a__ : Any = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : List[str] = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Union[Dict[str, EncodedInput], BatchEncoding] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : PaddingStrategy = PaddingStrategy.DO_NOT_PAD , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[bool] = None , ):
a__ : str = super()._pad(
encoded_inputs=lowerCamelCase__ , max_length=lowerCamelCase__ , padding_strategy=lowerCamelCase__ , pad_to_multiple_of=lowerCamelCase__ , return_attention_mask=lowerCamelCase__ , )
# Load from model defaults
if return_attention_mask is None:
a__ : Optional[int] = "attention_mask" in self.model_input_names
if return_attention_mask and "global_attention_mask" in encoded_inputs:
a__ : Tuple = encoded_inputs[self.model_input_names[0]]
# `global_attention_mask` need to have the same length as other (sequential) inputs.
a__ : Dict = len(encoded_inputs["global_attention_mask"] ) != len(lowerCamelCase__ )
if needs_to_be_padded:
a__ : Union[str, Any] = len(lowerCamelCase__ ) - len(encoded_inputs["global_attention_mask"] )
if self.padding_side == "right":
# Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend`
a__ : List[Any] = (
encoded_inputs["global_attention_mask"] + [-1] * difference
)
elif self.padding_side == "left":
a__ : Any = [-1] * difference + encoded_inputs[
"global_attention_mask"
]
else:
raise ValueError("Invalid padding strategy:" + str(self.padding_side ) )
return encoded_inputs
| 37 | 1 |
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] ):
a__ : Dict = ""
a__ : Optional[Any] = ""
a__ : Any = []
def _UpperCamelCase( self : Any , lowerCamelCase__ : int , lowerCamelCase__ : int ):
if m == -1:
return n + 1
elif n == -1:
return m + 1
elif self.dp[m][n] > -1:
return self.dp[m][n]
else:
if self.worda[m] == self.worda[n]:
a__ : List[str] = self.__min_dist_top_down_dp(m - 1 , n - 1 )
else:
a__ : Any = self.__min_dist_top_down_dp(lowerCamelCase__ , n - 1 )
a__ : Optional[Any] = self.__min_dist_top_down_dp(m - 1 , lowerCamelCase__ )
a__ : Optional[int] = self.__min_dist_top_down_dp(m - 1 , n - 1 )
a__ : str = 1 + min(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
return self.dp[m][n]
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str , lowerCamelCase__ : str ):
a__ : int = worda
a__ : List[str] = worda
a__ : Tuple = [[-1 for _ in range(len(lowerCamelCase__ ) )] for _ in range(len(lowerCamelCase__ ) )]
return self.__min_dist_top_down_dp(len(lowerCamelCase__ ) - 1 , len(lowerCamelCase__ ) - 1 )
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : str ):
a__ : List[Any] = worda
a__ : int = worda
a__ : int = len(lowerCamelCase__ )
a__ : Optional[Any] = len(lowerCamelCase__ )
a__ : Optional[int] = [[0 for _ in range(n + 1 )] for _ in range(m + 1 )]
for i in range(m + 1 ):
for j in range(n + 1 ):
if i == 0: # first string is empty
a__ : Dict = j
elif j == 0: # second string is empty
a__ : Tuple = i
elif worda[i - 1] == worda[j - 1]: # last characters are equal
a__ : Tuple = self.dp[i - 1][j - 1]
else:
a__ : Tuple = self.dp[i][j - 1]
a__ : Optional[Any] = self.dp[i - 1][j]
a__ : int = self.dp[i - 1][j - 1]
a__ : Optional[int] = 1 + min(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
return self.dp[m][n]
if __name__ == "__main__":
UpperCamelCase : Optional[Any] = EditDistance()
print("""****************** Testing Edit Distance DP Algorithm ******************""")
print()
UpperCamelCase : int = input("""Enter the first string: """).strip()
UpperCamelCase : Dict = input("""Enter the second string: """).strip()
print()
print(f"""The minimum edit distance is: {solver.min_dist_top_down(Sa, Sa)}""")
print(f"""The minimum edit distance is: {solver.min_dist_bottom_up(Sa, Sa)}""")
print()
print("""*************** End of Testing Edit Distance DP Algorithm ***************""")
| 37 |
import json
from typing import List, Optional, Tuple
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import logging
from .tokenization_roberta import RobertaTokenizer
UpperCamelCase : Any = logging.get_logger(__name__)
UpperCamelCase : Any = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Union[str, Any] = {
"""vocab_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/vocab.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/vocab.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/vocab.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/vocab.json""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/vocab.json""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/vocab.json"""
),
},
"""merges_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/merges.txt""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/merges.txt""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/merges.txt""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/merges.txt""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/merges.txt""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/merges.txt"""
),
},
"""tokenizer_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/tokenizer.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/tokenizer.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/tokenizer.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/tokenizer.json""",
"""roberta-base-openai-detector""": (
"""https://huggingface.co/roberta-base-openai-detector/resolve/main/tokenizer.json"""
),
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/tokenizer.json"""
),
},
}
UpperCamelCase : List[str] = {
"""roberta-base""": 512,
"""roberta-large""": 512,
"""roberta-large-mnli""": 512,
"""distilroberta-base""": 512,
"""roberta-base-openai-detector""": 512,
"""roberta-large-openai-detector""": 512,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = ['input_ids', 'attention_mask']
_lowercase = RobertaTokenizer
def __init__( self : List[str] , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[Any]=None , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]="replace" , lowerCamelCase__ : List[str]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Any="</s>" , lowerCamelCase__ : Any="<s>" , lowerCamelCase__ : int="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Tuple="<mask>" , lowerCamelCase__ : Any=False , lowerCamelCase__ : Dict=True , **lowerCamelCase__ : Optional[Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Any = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : int = add_prefix_space
a__ : Tuple = pre_tok_class(**lowerCamelCase__ )
a__ : str = add_prefix_space
a__ : Tuple = "post_processor"
a__ : Dict = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Tuple = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Tuple = tuple(state["sep"] )
if "cls" in state:
a__ : str = tuple(state["cls"] )
a__ : str = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : str = add_prefix_space
a__ : Any = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : int = trim_offsets
a__ : Dict = True
if changes_to_apply:
a__ : Union[str, Any] = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : str = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Tuple ):
a__ : List[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : List[str] = value
def _UpperCamelCase( self : Union[str, Any] , *lowerCamelCase__ : int , **lowerCamelCase__ : int ):
a__ : Optional[int] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Tuple , *lowerCamelCase__ : Dict , **lowerCamelCase__ : List[str] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : str , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : int = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[int]=None ):
a__ : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : Tuple = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
| 37 | 1 |
from __future__ import annotations
def UpperCamelCase_ ( __a , __a ) -> float:
a__ : Tuple = sorted(numsa + numsa )
a__, a__ : str = divmod(len(__a ) , 2 )
if mod == 1:
return all_numbers[div]
else:
return (all_numbers[div] + all_numbers[div - 1]) / 2
if __name__ == "__main__":
import doctest
doctest.testmod()
UpperCamelCase : Dict = [float(x) for x in input("""Enter the elements of first array: """).split()]
UpperCamelCase : Tuple = [float(x) for x in input("""Enter the elements of second array: """).split()]
print(f"""The median of two arrays is: {median_of_two_arrays(array_a, array_a)}""")
| 37 |
from statistics import mean, stdev
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : List[str] = min(__a )
a__ : str = max(__a )
# normalize data
return [round((x - x_min) / (x_max - x_min) , __a ) for x in data]
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : str = mean(__a )
a__ : List[str] = stdev(__a )
# standardize data
return [round((x - mu) / (sigma) , __a ) for x in data]
| 37 | 1 |
from __future__ import annotations
import inspect
import unittest
import numpy as np
from transformers import DeiTConfig
from transformers.testing_utils import require_tf, require_vision, slow
from transformers.utils import cached_property, is_tf_available, is_vision_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_tf_common import TFModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_tf_available():
import tensorflow as tf
from transformers import (
TFDeiTForImageClassification,
TFDeiTForImageClassificationWithTeacher,
TFDeiTForMaskedImageModeling,
TFDeiTModel,
)
from transformers.models.deit.modeling_tf_deit import TF_DEIT_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
from PIL import Image
from transformers import DeiTImageProcessor
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[Any]=13 , lowerCamelCase__ : Union[str, Any]=30 , lowerCamelCase__ : Dict=2 , lowerCamelCase__ : Optional[int]=3 , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : Union[str, Any]=True , lowerCamelCase__ : int=32 , lowerCamelCase__ : List[str]=2 , lowerCamelCase__ : List[Any]=4 , lowerCamelCase__ : Optional[Any]=37 , lowerCamelCase__ : Optional[int]="gelu" , lowerCamelCase__ : Tuple=0.1 , lowerCamelCase__ : List[Any]=0.1 , lowerCamelCase__ : List[str]=10 , lowerCamelCase__ : Union[str, Any]=0.02 , lowerCamelCase__ : int=3 , lowerCamelCase__ : Tuple=None , lowerCamelCase__ : Tuple=2 , ):
a__ : str = parent
a__ : Any = batch_size
a__ : Optional[int] = image_size
a__ : Tuple = patch_size
a__ : Union[str, Any] = num_channels
a__ : str = is_training
a__ : Dict = use_labels
a__ : List[str] = hidden_size
a__ : Union[str, Any] = num_hidden_layers
a__ : Tuple = num_attention_heads
a__ : Dict = intermediate_size
a__ : List[Any] = hidden_act
a__ : int = hidden_dropout_prob
a__ : Optional[Any] = attention_probs_dropout_prob
a__ : Dict = type_sequence_label_size
a__ : Optional[Any] = initializer_range
a__ : Tuple = scope
a__ : List[Any] = encoder_stride
# in DeiT, the seq length equals the number of patches + 2 (we add 2 for the [CLS] and distilation tokens)
a__ : Optional[Any] = (image_size // patch_size) ** 2
a__ : List[str] = num_patches + 2
def _UpperCamelCase( self : Optional[Any] ):
a__ : Optional[Any] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
a__ : List[str] = None
if self.use_labels:
a__ : int = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : str = self.get_config()
return config, pixel_values, labels
def _UpperCamelCase( self : Tuple ):
return DeiTConfig(
image_size=self.image_size , patch_size=self.patch_size , num_channels=self.num_channels , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , encoder_stride=self.encoder_stride , )
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Any , lowerCamelCase__ : int , lowerCamelCase__ : str ):
a__ : List[str] = TFDeiTModel(config=lowerCamelCase__ )
a__ : Optional[int] = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[int] ):
a__ : str = TFDeiTForMaskedImageModeling(config=lowerCamelCase__ )
a__ : Optional[int] = model(lowerCamelCase__ )
self.parent.assertEqual(
result.reconstruction.shape , (self.batch_size, self.num_channels, self.image_size, self.image_size) )
# test greyscale images
a__ : Dict = 1
a__ : Union[str, Any] = TFDeiTForMaskedImageModeling(lowerCamelCase__ )
a__ : List[Any] = floats_tensor([self.batch_size, 1, self.image_size, self.image_size] )
a__ : int = model(lowerCamelCase__ )
self.parent.assertEqual(result.reconstruction.shape , (self.batch_size, 1, self.image_size, self.image_size) )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : str ):
a__ : Optional[int] = self.type_sequence_label_size
a__ : Any = TFDeiTForImageClassification(lowerCamelCase__ )
a__ : List[Any] = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
# test greyscale images
a__ : List[str] = 1
a__ : str = TFDeiTForImageClassification(lowerCamelCase__ )
a__ : Any = floats_tensor([self.batch_size, 1, self.image_size, self.image_size] )
a__ : Union[str, Any] = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def _UpperCamelCase( self : List[str] ):
a__ : Optional[int] = self.prepare_config_and_inputs()
a__, a__, a__ : Union[str, Any] = config_and_inputs
a__ : List[Any] = {"pixel_values": pixel_values}
return config, inputs_dict
@require_tf
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(
TFDeiTModel,
TFDeiTForImageClassification,
TFDeiTForImageClassificationWithTeacher,
TFDeiTForMaskedImageModeling,
)
if is_tf_available()
else ()
)
_lowercase = (
{
'feature-extraction': TFDeiTModel,
'image-classification': (TFDeiTForImageClassification, TFDeiTForImageClassificationWithTeacher),
}
if is_tf_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Dict ):
a__ : Dict = TFDeiTModelTester(self )
a__ : Optional[int] = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[str] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="DeiT does not use inputs_embeds" )
def _UpperCamelCase( self : List[Any] ):
pass
def _UpperCamelCase( self : str ):
a__, a__ : int = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : List[str] = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (tf.keras.layers.Layer) )
a__ : Optional[int] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , tf.keras.layers.Dense ) )
def _UpperCamelCase( self : List[Any] ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : str = model_class(lowerCamelCase__ )
a__ : int = inspect.signature(model.call )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : List[Any] = [*signature.parameters.keys()]
a__ : Any = ["pixel_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
def _UpperCamelCase( self : str ):
a__ : Tuple = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_image_modeling(*lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : int , lowerCamelCase__ : List[str]=False ):
a__ : Union[str, Any] = super()._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
if return_labels:
if "labels" in inputs_dict and "labels" not in inspect.signature(model_class.call ).parameters:
del inputs_dict["labels"]
return inputs_dict
@slow
def _UpperCamelCase( self : int ):
for model_name in TF_DEIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Optional[Any] = TFDeiTModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Tuple:
a__ : Union[str, Any] = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
return image
@require_tf
@require_vision
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : Tuple ):
return (
DeiTImageProcessor.from_pretrained("facebook/deit-base-distilled-patch16-224" )
if is_vision_available()
else None
)
@slow
def _UpperCamelCase( self : Union[str, Any] ):
a__ : int = TFDeiTForImageClassificationWithTeacher.from_pretrained("facebook/deit-base-distilled-patch16-224" )
a__ : Any = self.default_image_processor
a__ : Any = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="tf" )
# forward pass
a__ : Dict = model(**lowerCamelCase__ )
# verify the logits
a__ : Any = tf.TensorShape((1, 1_000) )
self.assertEqual(outputs.logits.shape , lowerCamelCase__ )
a__ : Optional[Any] = tf.constant([-1.0266, 0.1912, -1.2861] )
self.assertTrue(np.allclose(outputs.logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
| 37 |
def UpperCamelCase_ ( __a = 50 ) -> int:
a__ : Tuple = [[0] * 3 for _ in range(length + 1 )]
for row_length in range(length + 1 ):
for tile_length in range(2 , 5 ):
for tile_start in range(row_length - tile_length + 1 ):
different_colour_ways_number[row_length][tile_length - 2] += (
different_colour_ways_number[row_length - tile_start - tile_length][
tile_length - 2
]
+ 1
)
return sum(different_colour_ways_number[length] )
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 | 1 |
from typing import TYPE_CHECKING
from ...utils import (
OptionalDependencyNotAvailable,
_LazyModule,
is_flax_available,
is_tf_available,
is_torch_available,
)
UpperCamelCase : Optional[int] = {
"""configuration_roberta_prelayernorm""": [
"""ROBERTA_PRELAYERNORM_PRETRAINED_CONFIG_ARCHIVE_MAP""",
"""RobertaPreLayerNormConfig""",
"""RobertaPreLayerNormOnnxConfig""",
],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
UpperCamelCase : Dict = [
"""ROBERTA_PRELAYERNORM_PRETRAINED_MODEL_ARCHIVE_LIST""",
"""RobertaPreLayerNormForCausalLM""",
"""RobertaPreLayerNormForMaskedLM""",
"""RobertaPreLayerNormForMultipleChoice""",
"""RobertaPreLayerNormForQuestionAnswering""",
"""RobertaPreLayerNormForSequenceClassification""",
"""RobertaPreLayerNormForTokenClassification""",
"""RobertaPreLayerNormModel""",
"""RobertaPreLayerNormPreTrainedModel""",
]
try:
if not is_tf_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
UpperCamelCase : List[Any] = [
"""TF_ROBERTA_PRELAYERNORM_PRETRAINED_MODEL_ARCHIVE_LIST""",
"""TFRobertaPreLayerNormForCausalLM""",
"""TFRobertaPreLayerNormForMaskedLM""",
"""TFRobertaPreLayerNormForMultipleChoice""",
"""TFRobertaPreLayerNormForQuestionAnswering""",
"""TFRobertaPreLayerNormForSequenceClassification""",
"""TFRobertaPreLayerNormForTokenClassification""",
"""TFRobertaPreLayerNormMainLayer""",
"""TFRobertaPreLayerNormModel""",
"""TFRobertaPreLayerNormPreTrainedModel""",
]
try:
if not is_flax_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
UpperCamelCase : Tuple = [
"""FlaxRobertaPreLayerNormForCausalLM""",
"""FlaxRobertaPreLayerNormForMaskedLM""",
"""FlaxRobertaPreLayerNormForMultipleChoice""",
"""FlaxRobertaPreLayerNormForQuestionAnswering""",
"""FlaxRobertaPreLayerNormForSequenceClassification""",
"""FlaxRobertaPreLayerNormForTokenClassification""",
"""FlaxRobertaPreLayerNormModel""",
"""FlaxRobertaPreLayerNormPreTrainedModel""",
]
if TYPE_CHECKING:
from .configuration_roberta_prelayernorm import (
ROBERTA_PRELAYERNORM_PRETRAINED_CONFIG_ARCHIVE_MAP,
RobertaPreLayerNormConfig,
RobertaPreLayerNormOnnxConfig,
)
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_roberta_prelayernorm import (
ROBERTA_PRELAYERNORM_PRETRAINED_MODEL_ARCHIVE_LIST,
RobertaPreLayerNormForCausalLM,
RobertaPreLayerNormForMaskedLM,
RobertaPreLayerNormForMultipleChoice,
RobertaPreLayerNormForQuestionAnswering,
RobertaPreLayerNormForSequenceClassification,
RobertaPreLayerNormForTokenClassification,
RobertaPreLayerNormModel,
RobertaPreLayerNormPreTrainedModel,
)
try:
if not is_tf_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_tf_roberta_prelayernorm import (
TF_ROBERTA_PRELAYERNORM_PRETRAINED_MODEL_ARCHIVE_LIST,
TFRobertaPreLayerNormForCausalLM,
TFRobertaPreLayerNormForMaskedLM,
TFRobertaPreLayerNormForMultipleChoice,
TFRobertaPreLayerNormForQuestionAnswering,
TFRobertaPreLayerNormForSequenceClassification,
TFRobertaPreLayerNormForTokenClassification,
TFRobertaPreLayerNormMainLayer,
TFRobertaPreLayerNormModel,
TFRobertaPreLayerNormPreTrainedModel,
)
try:
if not is_flax_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_flax_roberta_prelayernorm import (
FlaxRobertaPreLayerNormForCausalLM,
FlaxRobertaPreLayerNormForMaskedLM,
FlaxRobertaPreLayerNormForMultipleChoice,
FlaxRobertaPreLayerNormForQuestionAnswering,
FlaxRobertaPreLayerNormForSequenceClassification,
FlaxRobertaPreLayerNormForTokenClassification,
FlaxRobertaPreLayerNormModel,
FlaxRobertaPreLayerNormPreTrainedModel,
)
else:
import sys
UpperCamelCase : Optional[Any] = _LazyModule(__name__, globals()["""__file__"""], _import_structure, module_spec=__spec__)
| 37 |
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[str] ):
a__ : str = name
a__ : Optional[int] = value
a__ : Dict = weight
def __repr__( self : Union[str, Any] ):
return f'''{self.__class__.__name__}({self.name}, {self.value}, {self.weight})'''
def _UpperCamelCase( self : Dict ):
return self.value
def _UpperCamelCase( self : Optional[Any] ):
return self.name
def _UpperCamelCase( self : Optional[Any] ):
return self.weight
def _UpperCamelCase( self : Optional[int] ):
return self.value / self.weight
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = []
for i in range(len(__a ) ):
menu.append(Things(name[i] , value[i] , weight[i] ) )
return menu
def UpperCamelCase_ ( __a , __a , __a ) -> Union[str, Any]:
a__ : List[str] = sorted(__a , key=__a , reverse=__a )
a__ : List[Any] = []
a__, a__ : Union[str, Any] = 0.0, 0.0
for i in range(len(__a ) ):
if (total_cost + items_copy[i].get_weight()) <= max_cost:
result.append(items_copy[i] )
total_cost += items_copy[i].get_weight()
total_value += items_copy[i].get_value()
return (result, total_value)
def UpperCamelCase_ ( ) -> Union[str, Any]:
pass
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 | 1 |
def UpperCamelCase_ ( __a ) -> list:
a__ : Union[str, Any] = [0] * len(__a )
for i in range(1 , len(__a ) ):
# use last results for better performance - dynamic programming
a__ : Dict = prefix_result[i - 1]
while j > 0 and input_string[i] != input_string[j]:
a__ : List[Any] = prefix_result[j - 1]
if input_string[i] == input_string[j]:
j += 1
a__ : Any = j
return prefix_result
def UpperCamelCase_ ( __a ) -> int:
return max(prefix_function(__a ) )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 |
import multiprocessing
from typing import TYPE_CHECKING, Optional, Union
from .. import Dataset, Features, config
from ..formatting import query_table
from ..packaged_modules.sql.sql import Sql
from ..utils import logging
from .abc import AbstractDatasetInputStream
if TYPE_CHECKING:
import sqlitea
import sqlalchemy
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Dict , lowerCamelCase__ : Union[str, "sqlalchemy.sql.Selectable"] , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[Features] = None , lowerCamelCase__ : str = None , lowerCamelCase__ : bool = False , **lowerCamelCase__ : Optional[int] , ):
super().__init__(features=lowerCamelCase__ , cache_dir=lowerCamelCase__ , keep_in_memory=lowerCamelCase__ , **lowerCamelCase__ )
a__ : str = Sql(
cache_dir=lowerCamelCase__ , features=lowerCamelCase__ , sql=lowerCamelCase__ , con=lowerCamelCase__ , **lowerCamelCase__ , )
def _UpperCamelCase( self : Tuple ):
a__ : Optional[Any] = None
a__ : Dict = None
a__ : Union[str, Any] = None
a__ : Union[str, Any] = None
self.builder.download_and_prepare(
download_config=lowerCamelCase__ , download_mode=lowerCamelCase__ , verification_mode=lowerCamelCase__ , base_path=lowerCamelCase__ , )
# Build dataset for splits
a__ : List[str] = self.builder.as_dataset(
split="train" , verification_mode=lowerCamelCase__ , in_memory=self.keep_in_memory )
return dataset
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : Dataset , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[int] = None , **lowerCamelCase__ : Optional[Any] , ):
if num_proc is not None and num_proc <= 0:
raise ValueError(f'''num_proc {num_proc} must be an integer > 0.''' )
a__ : Any = dataset
a__ : str = name
a__ : Tuple = con
a__ : List[Any] = batch_size if batch_size else config.DEFAULT_MAX_BATCH_SIZE
a__ : Any = num_proc
a__ : Tuple = to_sql_kwargs
def _UpperCamelCase( self : List[Any] ):
a__ : Any = self.to_sql_kwargs.pop("sql" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("con" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("index" , lowerCamelCase__ )
a__ : int = self._write(index=lowerCamelCase__ , **self.to_sql_kwargs )
return written
def _UpperCamelCase( self : Any , lowerCamelCase__ : List[str] ):
a__, a__, a__ : Union[str, Any] = args
a__ : Any = {**to_sql_kwargs, "if_exists": "append"} if offset > 0 else to_sql_kwargs
a__ : Tuple = query_table(
table=self.dataset.data , key=slice(lowerCamelCase__ , offset + self.batch_size ) , indices=self.dataset._indices , )
a__ : str = batch.to_pandas()
a__ : List[Any] = df.to_sql(self.name , self.con , index=lowerCamelCase__ , **lowerCamelCase__ )
return num_rows or len(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : str = 0
if self.num_proc is None or self.num_proc == 1:
for offset in logging.tqdm(
range(0 , len(self.dataset ) , self.batch_size ) , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += self._batch_sql((offset, index, to_sql_kwargs) )
else:
a__, a__ : List[str] = len(self.dataset ), self.batch_size
with multiprocessing.Pool(self.num_proc ) as pool:
for num_rows in logging.tqdm(
pool.imap(
self._batch_sql , [(offset, index, to_sql_kwargs) for offset in range(0 , lowerCamelCase__ , lowerCamelCase__ )] , ) , total=(num_rows // batch_size) + 1 if num_rows % batch_size else num_rows // batch_size , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += num_rows
return written
| 37 | 1 |
import warnings
from ...processing_utils import ProcessorMixin
from ...tokenization_utils_base import BatchEncoding
class A__ ( A__ ):
"""simple docstring"""
_lowercase = ['image_processor', 'tokenizer']
_lowercase = 'CLIPImageProcessor'
_lowercase = ('CLIPTokenizer', 'CLIPTokenizerFast')
def __init__( self : Union[str, Any] , lowerCamelCase__ : Union[str, Any]=None , lowerCamelCase__ : Tuple=None , **lowerCamelCase__ : Optional[Any] ):
a__ : Optional[Any] = None
if "feature_extractor" in kwargs:
warnings.warn(
"The `feature_extractor` argument is deprecated and will be removed in v5, use `image_processor`"
" instead." , lowerCamelCase__ , )
a__ : List[Any] = kwargs.pop("feature_extractor" )
a__ : List[str] = image_processor if image_processor is not None else feature_extractor
if image_processor is None:
raise ValueError("You need to specify an `image_processor`." )
if tokenizer is None:
raise ValueError("You need to specify a `tokenizer`." )
super().__init__(lowerCamelCase__ , lowerCamelCase__ )
def __call__( self : List[Any] , lowerCamelCase__ : Dict=None , lowerCamelCase__ : Any=None , lowerCamelCase__ : str=None , **lowerCamelCase__ : Any ):
if text is None and images is None:
raise ValueError("You have to specify either text or images. Both cannot be none." )
if text is not None:
a__ : List[Any] = self.tokenizer(lowerCamelCase__ , return_tensors=lowerCamelCase__ , **lowerCamelCase__ )
if images is not None:
a__ : str = self.image_processor(lowerCamelCase__ , return_tensors=lowerCamelCase__ , **lowerCamelCase__ )
if text is not None and images is not None:
a__ : Any = image_features.pixel_values
return encoding
elif text is not None:
return encoding
else:
return BatchEncoding(data=dict(**lowerCamelCase__ ) , tensor_type=lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] , *lowerCamelCase__ : int , **lowerCamelCase__ : Optional[Any] ):
return self.tokenizer.batch_decode(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , *lowerCamelCase__ : str , **lowerCamelCase__ : Dict ):
return self.tokenizer.decode(*lowerCamelCase__ , **lowerCamelCase__ )
@property
def _UpperCamelCase( self : List[Any] ):
a__ : List[str] = self.tokenizer.model_input_names
a__ : Tuple = self.image_processor.model_input_names
return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names ) )
@property
def _UpperCamelCase( self : Optional[Any] ):
warnings.warn(
"`feature_extractor_class` is deprecated and will be removed in v5. Use `image_processor_class` instead." , lowerCamelCase__ , )
return self.image_processor_class
@property
def _UpperCamelCase( self : List[str] ):
warnings.warn(
"`feature_extractor` is deprecated and will be removed in v5. Use `image_processor` instead." , lowerCamelCase__ , )
return self.image_processor
| 37 |
import math
from datetime import datetime, timedelta
def UpperCamelCase_ ( __a ) -> datetime:
a__ : Union[str, Any] = year % 19
a__ : List[str] = year % 4
a__ : str = year % 7
a__ : Any = math.floor(year / 100 )
a__ : List[str] = math.floor((13 + 8 * leap_day_inhibits) / 25 )
a__ : Optional[int] = leap_day_inhibits / 4
a__ : Union[str, Any] = (
15 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 30
a__ : Dict = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
a__ : Any = (19 * metonic_cycle + secular_moon_shift) % 30
# PHM -> Paschal Full Moon
a__ : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 29 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 19 )
elif days_to_add == 28 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 18 )
else:
return datetime(__a , 3 , 22 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
UpperCamelCase : Tuple = """will be""" if year > datetime.now().year else """was"""
print(f"""Easter in {year} {tense} {gauss_easter(year)}""")
| 37 | 1 |
from typing import List, Optional, Union
import numpy as np
import torch
import torchaudio.compliance.kaldi as ta_kaldi
from ...feature_extraction_sequence_utils import SequenceFeatureExtractor
from ...feature_extraction_utils import BatchFeature
from ...utils import PaddingStrategy, TensorType, logging
UpperCamelCase : Tuple = logging.get_logger(__name__)
class A__ ( A__ ):
"""simple docstring"""
_lowercase = ['input_features', 'attention_mask']
def __init__( self : str , lowerCamelCase__ : List[str]=80 , lowerCamelCase__ : Any=16_000 , lowerCamelCase__ : int=80 , lowerCamelCase__ : Optional[Any]=0.0 , lowerCamelCase__ : Union[str, Any]=True , lowerCamelCase__ : Dict=True , lowerCamelCase__ : Tuple=True , **lowerCamelCase__ : Optional[int] , ):
super().__init__(feature_size=lowerCamelCase__ , sampling_rate=lowerCamelCase__ , padding_value=lowerCamelCase__ , **lowerCamelCase__ )
a__ : Dict = num_mel_bins
a__ : Optional[int] = do_ceptral_normalize
a__ : List[str] = normalize_means
a__ : Any = normalize_vars
a__ : Optional[Any] = True
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : np.ndarray , ):
a__ : Tuple = waveform * (2**15) # Kaldi compliance: 16-bit signed integers
a__ : Union[str, Any] = torch.from_numpy(lowerCamelCase__ ).unsqueeze(0 )
a__ : str = ta_kaldi.fbank(lowerCamelCase__ , num_mel_bins=self.num_mel_bins , sample_frequency=self.sampling_rate )
return features.numpy()
@staticmethod
def _UpperCamelCase( lowerCamelCase__ : np.ndarray , lowerCamelCase__ : int , lowerCamelCase__ : Optional[bool] = True , lowerCamelCase__ : Optional[bool] = True , lowerCamelCase__ : float = 0.0 , ):
# make sure we normalize float32 arrays
if normalize_means:
a__ : List[str] = x[:input_length].mean(axis=0 )
a__ : Dict = np.subtract(lowerCamelCase__ , lowerCamelCase__ )
if normalize_vars:
a__ : Any = x[:input_length].std(axis=0 )
a__ : List[str] = np.divide(lowerCamelCase__ , lowerCamelCase__ )
if input_length < x.shape[0]:
a__ : str = padding_value
# make sure array is in float32
a__ : List[Any] = x.astype(np.floataa )
return x
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[np.ndarray] , lowerCamelCase__ : Optional[np.ndarray] = None ):
a__ : List[str] = attention_mask.sum(-1 ) if attention_mask is not None else [x.shape[0] for x in input_features]
return [
self.utterance_cmvn(lowerCamelCase__ , lowerCamelCase__ , self.normalize_means , self.normalize_vars , self.padding_value )
for x, n in zip(lowerCamelCase__ , lowerCamelCase__ )
]
def __call__( self : str , lowerCamelCase__ : Union[np.ndarray, List[float], List[np.ndarray], List[List[float]]] , lowerCamelCase__ : Union[bool, str, PaddingStrategy] = False , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : bool = False , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[Union[str, TensorType]] = None , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[bool] = None , **lowerCamelCase__ : List[Any] , ):
if sampling_rate is not None:
if sampling_rate != self.sampling_rate:
raise ValueError(
f'''The model corresponding to this feature extractor: {self} was trained using a sampling rate of'''
f''' {self.sampling_rate}. Please make sure that the provided `raw_speech` input was sampled with'''
f''' {self.sampling_rate} and not {sampling_rate}.''' )
else:
logger.warning(
"It is strongly recommended to pass the `sampling_rate` argument to this function. "
"Failing to do so can result in silent errors that might be hard to debug." )
a__ : List[str] = isinstance(lowerCamelCase__ , np.ndarray ) and len(raw_speech.shape ) > 1
if is_batched_numpy and len(raw_speech.shape ) > 2:
raise ValueError(f'''Only mono-channel audio is supported for input to {self}''' )
a__ : Dict = is_batched_numpy or (
isinstance(lowerCamelCase__ , (list, tuple) ) and (isinstance(raw_speech[0] , (np.ndarray, tuple, list) ))
)
if is_batched:
a__ : str = [np.asarray(lowerCamelCase__ , dtype=np.floataa ) for speech in raw_speech]
elif not is_batched and not isinstance(lowerCamelCase__ , np.ndarray ):
a__ : Tuple = np.asarray(lowerCamelCase__ , dtype=np.floataa )
elif isinstance(lowerCamelCase__ , np.ndarray ) and raw_speech.dtype is np.dtype(np.floataa ):
a__ : str = raw_speech.astype(np.floataa )
# always return batch
if not is_batched:
a__ : List[str] = [raw_speech]
# extract fbank features
a__ : Any = [self._extract_fbank_features(lowerCamelCase__ ) for waveform in raw_speech]
# convert into correct format for padding
a__ : Any = BatchFeature({"input_features": features} )
a__ : Any = self.pad(
lowerCamelCase__ , padding=lowerCamelCase__ , max_length=lowerCamelCase__ , truncation=lowerCamelCase__ , pad_to_multiple_of=lowerCamelCase__ , return_attention_mask=lowerCamelCase__ , **lowerCamelCase__ , )
# make sure list is in array format
a__ : List[str] = padded_inputs.get("input_features" )
if isinstance(input_features[0] , lowerCamelCase__ ):
a__ : Optional[int] = [np.asarray(lowerCamelCase__ , dtype=np.floataa ) for feature in input_features]
a__ : Dict = padded_inputs.get("attention_mask" )
if attention_mask is not None:
a__ : Optional[int] = [np.asarray(lowerCamelCase__ , dtype=np.intaa ) for array in attention_mask]
# Utterance-level cepstral mean and variance normalization
if self.do_ceptral_normalize:
a__ : Optional[Any] = (
np.array(lowerCamelCase__ , dtype=np.intaa )
if self._get_padding_strategies(lowerCamelCase__ , max_length=lowerCamelCase__ ) is not PaddingStrategy.DO_NOT_PAD
else None
)
a__ : Optional[Any] = self.normalize(
padded_inputs["input_features"] , attention_mask=lowerCamelCase__ )
if return_tensors is not None:
a__ : Optional[Any] = padded_inputs.convert_to_tensors(lowerCamelCase__ )
return padded_inputs
| 37 |
import gc
import importlib.metadata
import tempfile
import unittest
from packaging import version
from transformers import (
AutoModel,
AutoModelForCausalLM,
AutoModelForSeqaSeqLM,
AutoModelForSequenceClassification,
AutoTokenizer,
BitsAndBytesConfig,
pipeline,
)
from transformers.testing_utils import (
is_torch_available,
require_accelerate,
require_bitsandbytes,
require_torch,
require_torch_gpu,
require_torch_multi_gpu,
slow,
)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
if model.config.model_type == "gpt2":
return model.transformer.h[0].mlp.c_fc
return model.transformer.h[0].mlp.dense_ah_to_h
if is_torch_available():
import torch
import torch.nn as nn
class A__ ( nn.Module ):
"""simple docstring"""
def __init__( self : List[str] , lowerCamelCase__ : nn.Module , lowerCamelCase__ : int ):
super().__init__()
a__ : int = module
a__ : Any = nn.Sequential(
nn.Linear(module.in_features , lowerCamelCase__ , bias=lowerCamelCase__ ) , nn.Linear(lowerCamelCase__ , module.out_features , bias=lowerCamelCase__ ) , )
a__ : Tuple = (2.0 / (5 * min(module.in_features , module.out_features ))) ** 0.5
nn.init.normal_(self.adapter[0].weight , std=lowerCamelCase__ )
nn.init.zeros_(self.adapter[1].weight )
self.adapter.to(module.weight.device )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Optional[int] , *lowerCamelCase__ : int , **lowerCamelCase__ : Dict ):
return self.module(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ ) + self.adapter(lowerCamelCase__ )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
_lowercase = 'bigscience/bloom-1b7'
# Constant values
_lowercase = 2.1_09_65_95_52_69_25_74
_lowercase = 'Hello my name is'
_lowercase = set()
EXPECTED_OUTPUTS.add('Hello my name is John and I am a professional photographer. I' )
EXPECTED_OUTPUTS.add('Hello my name is John.\nI am a friend of your father.\n' )
EXPECTED_OUTPUTS.add('Hello my name is John Doe, I am a student at the University' )
_lowercase = 1_0
def _UpperCamelCase( self : Dict ):
# Models and tokenizer
a__ : List[str] = AutoTokenizer.from_pretrained(self.model_name )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Union[str, Any] ):
super().setUp()
# Models and tokenizer
a__ : List[Any] = AutoModelForCausalLM.from_pretrained(
self.model_name , torch_dtype=torch.floataa , device_map="auto" )
a__ : Union[str, Any] = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.model_fpaa
del self.model_abit
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : List[Any] ):
a__ : str = self.model_abit.config
self.assertTrue(hasattr(lowerCamelCase__ , "quantization_config" ) )
a__ : Optional[Any] = config.to_dict()
a__ : int = config.to_diff_dict()
a__ : List[str] = config.to_json_string()
def _UpperCamelCase( self : int ):
from bitsandbytes.nn import Paramsabit
a__ : List[Any] = self.model_fpaa.get_memory_footprint()
a__ : str = self.model_abit.get_memory_footprint()
self.assertAlmostEqual(mem_fpaa / mem_abit , self.EXPECTED_RELATIVE_DIFFERENCE )
a__ : Optional[Any] = get_some_linear_layer(self.model_abit )
self.assertTrue(linear.weight.__class__ == Paramsabit )
def _UpperCamelCase( self : Tuple ):
from transformers import TaPreTrainedModel
self.model_fpaa.get_memory_footprint()
self.model_abit.get_memory_footprint()
for name, module in self.model_abit.named_modules():
if isinstance(lowerCamelCase__ , torch.nn.Linear ):
if name not in ["lm_head"] + TaPreTrainedModel._keep_in_fpaa_modules:
# 4-bit parameters are packed in uint8 variables
self.assertTrue(module.weight.dtype == torch.uinta )
def _UpperCamelCase( self : str ):
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Tuple = self.model_abit.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[Any] = BitsAndBytesConfig()
a__ : Optional[int] = True
a__ : int = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , device_map="auto" )
a__ : str = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : int = model_abit_from_config.generate(
input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : Dict ):
with self.assertRaises(lowerCamelCase__ ), tempfile.TemporaryDirectory() as tmpdirname:
self.model_abit.save_pretrained(lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : int = BitsAndBytesConfig()
with self.assertRaises(lowerCamelCase__ ):
a__ : Dict = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , load_in_abit=lowerCamelCase__ , device_map="auto" , bnb_abit_quant_type="nf4" , )
def _UpperCamelCase( self : int ):
with self.assertRaises(lowerCamelCase__ ):
# Tries with `str`
self.model_abit.to("cpu" )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `dtype``
self.model_abit.to(torch.floataa )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.to(torch.device("cuda:0" ) )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.float()
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.half()
# Test if we did not break anything
a__ : int = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Any = self.model_fpaa.to(torch.floataa )
a__ : List[Any] = self.model_fpaa.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.to("cpu" )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.half()
# Check this does not throw an error
a__ : Dict = self.model_fpaa.float()
def _UpperCamelCase( self : Dict ):
a__ : List[str] = AutoModelForSeqaSeqLM.from_pretrained("t5-small" , load_in_abit=lowerCamelCase__ , device_map="auto" )
self.assertTrue(model.decoder.block[0].layer[2].DenseReluDense.wo.weight.dtype == torch.floataa )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
@classmethod
def _UpperCamelCase( cls : str ):
a__ : Dict = "t5-small"
a__ : List[Any] = "google/flan-t5-small" # flan-t5 uses dense-act instead of dense-relu-dense
a__ : int = AutoTokenizer.from_pretrained(cls.model_name )
a__ : str = "Translate in German: Hello, my dog is cute"
def _UpperCamelCase( self : Optional[int] ):
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Optional[int] ):
from transformers import TaForConditionalGeneration
a__ : List[Any] = TaForConditionalGeneration._keep_in_fpaa_modules
a__ : Optional[Any] = None
# test with `t5-small`
a__ : Any = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : Dict = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Any = model.generate(**lowerCamelCase__ )
a__ : Union[str, Any] = modules
def _UpperCamelCase( self : List[Any] ):
import bitsandbytes as bnb
from transformers import TaForConditionalGeneration
# test with `t5-small`
a__ : List[str] = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# there was a bug with decoders - this test checks that it is fixed
self.assertTrue(isinstance(model.decoder.block[0].layer[0].SelfAttention.q , bnb.nn.Linearabit ) )
a__ : Union[str, Any] = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : int = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Any = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Optional[int] = model.generate(**lowerCamelCase__ )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : List[str] ):
super().setUp()
# model_name
a__ : Union[str, Any] = "bigscience/bloom-560m"
a__ : Union[str, Any] = "t5-small"
# Different types of model
a__ : int = AutoModel.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Sequence classification model
a__ : Dict = AutoModelForSequenceClassification.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# CausalLM model
a__ : str = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Seq2seq model
a__ : Dict = AutoModelForSeqaSeqLM.from_pretrained(
self.seq_to_seq_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.base_model
del self.sequence_model
del self.model_abit
del self.seq_to_seq_model
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Union[str, Any] ):
from bitsandbytes.nn import Paramsabit
self.assertTrue(self.base_model.h[-1].mlp.dense_ah_to_h.weight.__class__ == Paramsabit )
# Other heads should be nn.Parameter
self.assertTrue(self.model_abit.lm_head.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.sequence_model.score.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.seq_to_seq_model.lm_head.weight.__class__ == torch.nn.Parameter )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
super().setUp()
def _UpperCamelCase( self : int ):
del self.pipe
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Tuple ):
a__ : int = pipeline(
"text-generation" , model=self.model_name , model_kwargs={"device_map": "auto", "load_in_4bit": True, "torch_dtype": torch.floataa} , max_new_tokens=self.MAX_NEW_TOKENS , )
# Real second forward pass
a__ : Tuple = self.pipe(self.input_text )
self.assertIn(pipeline_output[0]["generated_text"] , self.EXPECTED_OUTPUTS )
@require_torch_multi_gpu
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Tuple ):
super().setUp()
def _UpperCamelCase( self : List[Any] ):
a__ : str = AutoModelForCausalLM.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="balanced" )
# Check correct device map
self.assertEqual(set(model_parallel.hf_device_map.values() ) , {0, 1} )
# Check that inference pass works on the model
a__ : List[Any] = self.tokenizer(self.input_text , return_tensors="pt" )
# Second real batch
a__ : List[Any] = model_parallel.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_parallel[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
a__ : Any = "facebook/opt-350m"
super().setUp()
def _UpperCamelCase( self : int ):
if version.parse(importlib.metadata.version("bitsandbytes" ) ) < version.parse("0.37.0" ):
return
# Step 1: freeze all parameters
a__ : Tuple = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ )
self.assertEqual(set(model.hf_device_map.values() ) , {torch.cuda.current_device()} )
for param in model.parameters():
a__ : Any = False # freeze the model - train adapters later
if param.ndim == 1:
# cast the small parameters (e.g. layernorm) to fp32 for stability
a__ : Tuple = param.data.to(torch.floataa )
# Step 2: add adapters
for _, module in model.named_modules():
if "OPTAttention" in repr(type(lowerCamelCase__ ) ):
a__ : Dict = LoRALayer(module.q_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.k_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.v_proj , rank=16 )
# Step 3: dummy batch
a__ : Dict = self.tokenizer("Test batch " , return_tensors="pt" ).to(0 )
# Step 4: Check if the gradient is not None
with torch.cuda.amp.autocast():
a__ : Optional[Any] = model.forward(**lowerCamelCase__ )
out.logits.norm().backward()
for module in model.modules():
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertTrue(module.adapter[1].weight.grad is not None )
self.assertTrue(module.adapter[1].weight.grad.norm().item() > 0 )
elif isinstance(lowerCamelCase__ , nn.Embedding ):
self.assertTrue(module.weight.grad is None )
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gpt2-xl'
_lowercase = 3.31_91_85_48_54_15_21_87
| 37 | 1 |
from math import sqrt
def UpperCamelCase_ ( __a = 1_000_000 ) -> int:
a__ : int = 0
a__ : int = 0
a__ : int
while num_cuboids <= limit:
max_cuboid_size += 1
for sum_shortest_sides in range(2 , 2 * max_cuboid_size + 1 ):
if sqrt(sum_shortest_sides**2 + max_cuboid_size**2 ).is_integer():
num_cuboids += (
min(__a , sum_shortest_sides // 2 )
- max(1 , sum_shortest_sides - max_cuboid_size )
+ 1
)
return max_cuboid_size
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 |
import inspect
import unittest
from datasets import load_dataset
from packaging import version
from transformers import BeitConfig
from transformers.models.auto import get_values
from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_vision_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, _config_zero_init, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import (
MODEL_MAPPING,
BeitForImageClassification,
BeitForMaskedImageModeling,
BeitForSemanticSegmentation,
BeitModel,
)
from transformers.models.beit.modeling_beit import BEIT_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
import PIL
from PIL import Image
from transformers import BeitImageProcessor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Optional[int]=100 , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[int]=30 , lowerCamelCase__ : Union[str, Any]=2 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : List[Any]=True , lowerCamelCase__ : Tuple=True , lowerCamelCase__ : int=32 , lowerCamelCase__ : Union[str, Any]=4 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Union[str, Any]=37 , lowerCamelCase__ : List[Any]="gelu" , lowerCamelCase__ : List[Any]=0.1 , lowerCamelCase__ : int=0.1 , lowerCamelCase__ : Union[str, Any]=10 , lowerCamelCase__ : str=0.02 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]=[0, 1, 2, 3] , ):
a__ : Dict = parent
a__ : Dict = 100
a__ : Optional[int] = batch_size
a__ : Union[str, Any] = image_size
a__ : Any = patch_size
a__ : Optional[Any] = num_channels
a__ : int = is_training
a__ : List[str] = use_labels
a__ : Optional[Any] = hidden_size
a__ : List[Any] = num_hidden_layers
a__ : str = num_attention_heads
a__ : str = intermediate_size
a__ : int = hidden_act
a__ : List[Any] = hidden_dropout_prob
a__ : Dict = attention_probs_dropout_prob
a__ : Union[str, Any] = type_sequence_label_size
a__ : Optional[Any] = initializer_range
a__ : List[str] = scope
a__ : int = out_indices
a__ : List[str] = num_labels
# in BeiT, the seq length equals the number of patches + 1 (we add 1 for the [CLS] token)
a__ : Optional[int] = (image_size // patch_size) ** 2
a__ : Union[str, Any] = num_patches + 1
def _UpperCamelCase( self : int ):
a__ : Optional[int] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
a__ : Optional[Any] = None
a__ : Tuple = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : Dict = ids_tensor([self.batch_size, self.image_size, self.image_size] , self.num_labels )
a__ : Optional[int] = self.get_config()
return config, pixel_values, labels, pixel_labels
def _UpperCamelCase( self : Tuple ):
return BeitConfig(
vocab_size=self.vocab_size , image_size=self.image_size , patch_size=self.patch_size , num_channels=self.num_channels , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , out_indices=self.out_indices , )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : int , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any ):
a__ : str = BeitModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[str] , lowerCamelCase__ : Any , lowerCamelCase__ : List[str] , lowerCamelCase__ : Tuple ):
a__ : int = BeitForMaskedImageModeling(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[Any] = model(lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length - 1, self.vocab_size) )
def _UpperCamelCase( self : str , lowerCamelCase__ : Any , lowerCamelCase__ : str , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Any ):
a__ : List[str] = self.type_sequence_label_size
a__ : Optional[Any] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
# test greyscale images
a__ : Optional[Any] = 1
a__ : List[str] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = floats_tensor([self.batch_size, 1, self.image_size, self.image_size] )
a__ : Union[str, Any] = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def _UpperCamelCase( self : Any , lowerCamelCase__ : str , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
a__ : int = self.num_labels
a__ : List[str] = BeitForSemanticSegmentation(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Tuple = model(lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
def _UpperCamelCase( self : Optional[int] ):
a__ : Any = self.prepare_config_and_inputs()
a__, a__, a__, a__ : Union[str, Any] = config_and_inputs
a__ : Dict = {"pixel_values": pixel_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(BeitModel, BeitForImageClassification, BeitForMaskedImageModeling, BeitForSemanticSegmentation)
if is_torch_available()
else ()
)
_lowercase = (
{
'feature-extraction': BeitModel,
'image-classification': BeitForImageClassification,
'image-segmentation': BeitForSemanticSegmentation,
}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Any ):
a__ : int = BeitModelTester(self )
a__ : Optional[Any] = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[Any] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="BEiT does not use inputs_embeds" )
def _UpperCamelCase( self : str ):
pass
@require_torch_multi_gpu
@unittest.skip(reason="BEiT has some layers using `add_module` which doesn't work well with `nn.DataParallel`" )
def _UpperCamelCase( self : Dict ):
pass
def _UpperCamelCase( self : Optional[Any] ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : List[str] = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Optional[int] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : str ):
a__, a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : int = model_class(lowerCamelCase__ )
a__ : str = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Any = ["pixel_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
def _UpperCamelCase( self : int ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_lm(*lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] ):
a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_semantic_segmentation(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
if not self.model_tester.is_training:
return
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : str = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]:
continue
a__ : List[str] = model_class(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.train()
a__ : Any = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : Tuple = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : Tuple ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
if not self.model_tester.is_training:
return
a__ : List[Any] = False
a__ : List[str] = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if (
model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]
or not model_class.supports_gradient_checkpointing
):
continue
a__ : Optional[Any] = model_class(lowerCamelCase__ )
model.gradient_checkpointing_enable()
model.to(lowerCamelCase__ )
model.train()
a__ : Union[str, Any] = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : int = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : List[str] ):
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : Dict = _config_zero_init(lowerCamelCase__ )
for model_class in self.all_model_classes:
a__ : str = model_class(config=lowerCamelCase__ )
for name, param in model.named_parameters():
# we skip lambda parameters as these require special initial values
# determined by config.layer_scale_init_value
if "lambda" in name:
continue
if param.requires_grad:
self.assertIn(
((param.data.mean() * 1E9).round() / 1E9).item() , [0.0, 1.0] , msg=f'''Parameter {name} of model {model_class} seems not properly initialized''' , )
@slow
def _UpperCamelCase( self : Optional[int] ):
for model_name in BEIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Tuple = BeitModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : List[Any] = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
return image
@require_torch
@require_vision
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : Optional[int] ):
return BeitImageProcessor.from_pretrained("microsoft/beit-base-patch16-224" ) if is_vision_available() else None
@slow
def _UpperCamelCase( self : str ):
a__ : int = BeitForMaskedImageModeling.from_pretrained("microsoft/beit-base-patch16-224-pt22k" ).to(lowerCamelCase__ )
a__ : Optional[Any] = self.default_image_processor
a__ : Dict = prepare_img()
a__ : Optional[int] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).pixel_values.to(lowerCamelCase__ )
# prepare bool_masked_pos
a__ : Optional[Any] = torch.ones((1, 196) , dtype=torch.bool ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(pixel_values=lowerCamelCase__ , bool_masked_pos=lowerCamelCase__ )
a__ : Tuple = outputs.logits
# verify the logits
a__ : List[str] = torch.Size((1, 196, 8_192) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[int] = torch.tensor(
[[-3.2437, 0.5072, -13.9174], [-3.2456, 0.4948, -13.9401], [-3.2033, 0.5121, -13.8550]] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[bool_masked_pos][:3, :3] , lowerCamelCase__ , atol=1E-2 ) )
@slow
def _UpperCamelCase( self : Dict ):
a__ : str = BeitForImageClassification.from_pretrained("microsoft/beit-base-patch16-224" ).to(lowerCamelCase__ )
a__ : int = self.default_image_processor
a__ : List[Any] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Union[str, Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 1_000) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = torch.tensor([-1.2385, -1.0987, -1.0108] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Tuple = 281
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : Any ):
a__ : Dict = BeitForImageClassification.from_pretrained("microsoft/beit-large-patch16-224-pt22k-ft22k" ).to(
lowerCamelCase__ )
a__ : str = self.default_image_processor
a__ : List[str] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Dict = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Optional[int] = torch.Size((1, 21_841) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[Any] = torch.tensor([1.6881, -0.2787, 0.5901] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Optional[Any] = 2_396
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
a__ : Optional[Any] = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : Tuple = model.to(lowerCamelCase__ )
a__ : List[Any] = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Tuple = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : Union[str, Any] = Image.open(ds[0]["file"] )
a__ : List[Any] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Optional[Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Tuple = torch.Size((1, 150, 160, 160) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = version.parse(PIL.__version__ ) < version.parse("9.0.0" )
if is_pillow_less_than_a:
a__ : Dict = torch.tensor(
[
[[-4.9225, -2.3954, -3.0522], [-2.8822, -1.0046, -1.7561], [-2.9549, -1.3228, -2.1347]],
[[-5.8168, -3.4129, -4.0778], [-3.8651, -2.2214, -3.0277], [-3.8356, -2.4643, -3.3535]],
[[-0.0078, 3.9952, 4.0754], [2.9856, 4.6944, 5.0035], [3.2413, 4.7813, 4.9969]],
] , device=lowerCamelCase__ , )
else:
a__ : Dict = torch.tensor(
[
[[-4.8960, -2.3688, -3.0355], [-2.8478, -0.9836, -1.7418], [-2.9449, -1.3332, -2.1456]],
[[-5.8081, -3.4124, -4.1006], [-3.8561, -2.2081, -3.0323], [-3.8365, -2.4601, -3.3669]],
[[-0.0309, 3.9868, 4.0540], [2.9640, 4.6877, 4.9976], [3.2081, 4.7690, 4.9942]],
] , device=lowerCamelCase__ , )
self.assertTrue(torch.allclose(logits[0, :3, :3, :3] , lowerCamelCase__ , atol=1E-4 ) )
@slow
def _UpperCamelCase( self : Tuple ):
a__ : str = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : List[Any] = model.to(lowerCamelCase__ )
a__ : int = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Optional[int] = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : str = Image.open(ds[0]["file"] )
a__ : str = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : List[Any] = model(**lowerCamelCase__ )
a__ : Any = outputs.logits.detach().cpu()
a__ : List[Any] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ , target_sizes=[(500, 300)] )
a__ : Optional[int] = torch.Size((500, 300) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
a__ : List[str] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ )
a__ : Any = torch.Size((160, 160) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
| 37 | 1 |
import json
from typing import List, Optional, Tuple
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import logging
from .tokenization_roberta import RobertaTokenizer
UpperCamelCase : Any = logging.get_logger(__name__)
UpperCamelCase : Any = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Union[str, Any] = {
"""vocab_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/vocab.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/vocab.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/vocab.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/vocab.json""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/vocab.json""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/vocab.json"""
),
},
"""merges_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/merges.txt""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/merges.txt""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/merges.txt""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/merges.txt""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/merges.txt""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/merges.txt"""
),
},
"""tokenizer_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/tokenizer.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/tokenizer.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/tokenizer.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/tokenizer.json""",
"""roberta-base-openai-detector""": (
"""https://huggingface.co/roberta-base-openai-detector/resolve/main/tokenizer.json"""
),
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/tokenizer.json"""
),
},
}
UpperCamelCase : List[str] = {
"""roberta-base""": 512,
"""roberta-large""": 512,
"""roberta-large-mnli""": 512,
"""distilroberta-base""": 512,
"""roberta-base-openai-detector""": 512,
"""roberta-large-openai-detector""": 512,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = ['input_ids', 'attention_mask']
_lowercase = RobertaTokenizer
def __init__( self : List[str] , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[Any]=None , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]="replace" , lowerCamelCase__ : List[str]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Any="</s>" , lowerCamelCase__ : Any="<s>" , lowerCamelCase__ : int="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Tuple="<mask>" , lowerCamelCase__ : Any=False , lowerCamelCase__ : Dict=True , **lowerCamelCase__ : Optional[Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Any = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : int = add_prefix_space
a__ : Tuple = pre_tok_class(**lowerCamelCase__ )
a__ : str = add_prefix_space
a__ : Tuple = "post_processor"
a__ : Dict = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Tuple = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Tuple = tuple(state["sep"] )
if "cls" in state:
a__ : str = tuple(state["cls"] )
a__ : str = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : str = add_prefix_space
a__ : Any = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : int = trim_offsets
a__ : Dict = True
if changes_to_apply:
a__ : Union[str, Any] = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : str = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Tuple ):
a__ : List[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : List[str] = value
def _UpperCamelCase( self : Union[str, Any] , *lowerCamelCase__ : int , **lowerCamelCase__ : int ):
a__ : Optional[int] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Tuple , *lowerCamelCase__ : Dict , **lowerCamelCase__ : List[str] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : str , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : int = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[int]=None ):
a__ : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : Tuple = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
| 37 |
import re
import jax.numpy as jnp
from flax.traverse_util import flatten_dict, unflatten_dict
from jax.random import PRNGKey
from ..utils import logging
UpperCamelCase : Dict = logging.get_logger(__name__)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
a__ : Tuple = R"\w+[.]\d+"
a__ : List[Any] = re.findall(__a , __a )
for pat in pats:
a__ : Union[str, Any] = key.replace(__a , "_".join(pat.split("." ) ) )
return key
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : List[str] = pt_tuple_key[:-1] + ("scale",)
if (
any("norm" in str_ for str_ in pt_tuple_key )
and (pt_tuple_key[-1] == "bias")
and (pt_tuple_key[:-1] + ("bias",) not in random_flax_state_dict)
and (pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict)
):
a__ : Any = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
elif pt_tuple_key[-1] in ["weight", "gamma"] and pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict:
a__ : Optional[Any] = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
# embedding
if pt_tuple_key[-1] == "weight" and pt_tuple_key[:-1] + ("embedding",) in random_flax_state_dict:
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("embedding",)
return renamed_pt_tuple_key, pt_tensor
# conv layer
a__ : List[str] = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight" and pt_tensor.ndim == 4:
a__ : str = pt_tensor.transpose(2 , 3 , 1 , 0 )
return renamed_pt_tuple_key, pt_tensor
# linear layer
a__ : Tuple = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight":
a__ : Tuple = pt_tensor.T
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm weight
a__ : Optional[Any] = pt_tuple_key[:-1] + ("weight",)
if pt_tuple_key[-1] == "gamma":
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm bias
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("bias",)
if pt_tuple_key[-1] == "beta":
return renamed_pt_tuple_key, pt_tensor
return pt_tuple_key, pt_tensor
def UpperCamelCase_ ( __a , __a , __a=42 ) -> str:
# Step 1: Convert pytorch tensor to numpy
a__ : Optional[int] = {k: v.numpy() for k, v in pt_state_dict.items()}
# Step 2: Since the model is stateless, get random Flax params
a__ : Tuple = flax_model.init_weights(PRNGKey(__a ) )
a__ : Optional[Any] = flatten_dict(__a )
a__ : Union[str, Any] = {}
# Need to change some parameters name to match Flax names
for pt_key, pt_tensor in pt_state_dict.items():
a__ : Optional[int] = rename_key(__a )
a__ : Optional[int] = tuple(renamed_pt_key.split("." ) )
# Correctly rename weight parameters
a__, a__ : Union[str, Any] = rename_key_and_reshape_tensor(__a , __a , __a )
if flax_key in random_flax_state_dict:
if flax_tensor.shape != random_flax_state_dict[flax_key].shape:
raise ValueError(
f'''PyTorch checkpoint seems to be incorrect. Weight {pt_key} was expected to be of shape '''
f'''{random_flax_state_dict[flax_key].shape}, but is {flax_tensor.shape}.''' )
# also add unexpected weight so that warning is thrown
a__ : str = jnp.asarray(__a )
return unflatten_dict(__a )
| 37 | 1 |
import argparse
import collections
import json
import os
import re
import string
import sys
import numpy as np
UpperCamelCase : List[str] = re.compile(r"""\b(a|an|the)\b""", re.UNICODE)
UpperCamelCase : Union[str, Any] = None
def UpperCamelCase_ ( ) -> List[str]:
a__ : List[Any] = argparse.ArgumentParser("Official evaluation script for SQuAD version 2.0." )
parser.add_argument("data_file" , metavar="data.json" , help="Input data JSON file." )
parser.add_argument("pred_file" , metavar="pred.json" , help="Model predictions." )
parser.add_argument(
"--out-file" , "-o" , metavar="eval.json" , help="Write accuracy metrics to file (default is stdout)." )
parser.add_argument(
"--na-prob-file" , "-n" , metavar="na_prob.json" , help="Model estimates of probability of no answer." )
parser.add_argument(
"--na-prob-thresh" , "-t" , type=__a , default=1.0 , help="Predict \"\" if no-answer probability exceeds this (default = 1.0)." , )
parser.add_argument(
"--out-image-dir" , "-p" , metavar="out_images" , default=__a , help="Save precision-recall curves to directory." )
parser.add_argument("--verbose" , "-v" , action="store_true" )
if len(sys.argv ) == 1:
parser.print_help()
sys.exit(1 )
return parser.parse_args()
def UpperCamelCase_ ( __a ) -> str:
a__ : Optional[Any] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : Dict = bool(qa["answers"]["text"] )
return qid_to_has_ans
def UpperCamelCase_ ( __a ) -> List[Any]:
def remove_articles(__a ):
return ARTICLES_REGEX.sub(" " , __a )
def white_space_fix(__a ):
return " ".join(text.split() )
def remove_punc(__a ):
a__ : Union[str, Any] = set(string.punctuation )
return "".join(ch for ch in text if ch not in exclude )
def lower(__a ):
return text.lower()
return white_space_fix(remove_articles(remove_punc(lower(__a ) ) ) )
def UpperCamelCase_ ( __a ) -> Dict:
if not s:
return []
return normalize_answer(__a ).split()
def UpperCamelCase_ ( __a , __a ) -> str:
return int(normalize_answer(__a ) == normalize_answer(__a ) )
def UpperCamelCase_ ( __a , __a ) -> Dict:
a__ : int = get_tokens(__a )
a__ : Optional[Any] = get_tokens(__a )
a__ : Any = collections.Counter(__a ) & collections.Counter(__a )
a__ : Dict = sum(common.values() )
if len(__a ) == 0 or len(__a ) == 0:
# If either is no-answer, then F1 is 1 if they agree, 0 otherwise
return int(gold_toks == pred_toks )
if num_same == 0:
return 0
a__ : Tuple = 1.0 * num_same / len(__a )
a__ : str = 1.0 * num_same / len(__a )
a__ : str = (2 * precision * recall) / (precision + recall)
return fa
def UpperCamelCase_ ( __a , __a ) -> int:
a__ : List[str] = {}
a__ : Optional[int] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : List[Any] = qa["id"]
a__ : Dict = [t for t in qa["answers"]["text"] if normalize_answer(__a )]
if not gold_answers:
# For unanswerable questions, only correct answer is empty string
a__ : Tuple = [""]
if qid not in preds:
print(f'''Missing prediction for {qid}''' )
continue
a__ : Tuple = preds[qid]
# Take max over all gold answers
a__ : Optional[int] = max(compute_exact(__a , __a ) for a in gold_answers )
a__ : str = max(compute_fa(__a , __a ) for a in gold_answers )
return exact_scores, fa_scores
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
a__ : Optional[Any] = {}
for qid, s in scores.items():
a__ : Dict = na_probs[qid] > na_prob_thresh
if pred_na:
a__ : Dict = float(not qid_to_has_ans[qid] )
else:
a__ : Optional[Any] = s
return new_scores
def UpperCamelCase_ ( __a , __a , __a=None ) -> Tuple:
if not qid_list:
a__ : Union[str, Any] = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores.values() ) / total),
("f1", 100.0 * sum(fa_scores.values() ) / total),
("total", total),
] )
else:
a__ : int = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores[k] for k in qid_list ) / total),
("f1", 100.0 * sum(fa_scores[k] for k in qid_list ) / total),
("total", total),
] )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
for k in new_eval:
a__ : Optional[Any] = new_eval[k]
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
plt.step(__a , __a , color="b" , alpha=0.2 , where="post" )
plt.fill_between(__a , __a , step="post" , alpha=0.2 , color="b" )
plt.xlabel("Recall" )
plt.ylabel("Precision" )
plt.xlim([0.0, 1.05] )
plt.ylim([0.0, 1.05] )
plt.title(__a )
plt.savefig(__a )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a , __a=None , __a=None ) -> Dict:
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
a__ : Any = 0.0
a__ : Optional[int] = 1.0
a__ : Optional[int] = 0.0
a__ : Any = [1.0]
a__ : Tuple = [0.0]
a__ : List[str] = 0.0
for i, qid in enumerate(__a ):
if qid_to_has_ans[qid]:
true_pos += scores[qid]
a__ : Any = true_pos / float(i + 1 )
a__ : int = true_pos / float(__a )
if i == len(__a ) - 1 or na_probs[qid] != na_probs[qid_list[i + 1]]:
# i.e., if we can put a threshold after this point
avg_prec += cur_p * (cur_r - recalls[-1])
precisions.append(__a )
recalls.append(__a )
if out_image:
plot_pr_curve(__a , __a , __a , __a )
return {"ap": 100.0 * avg_prec}
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> str:
if out_image_dir and not os.path.exists(__a ):
os.makedirs(__a )
a__ : Optional[int] = sum(1 for v in qid_to_has_ans.values() if v )
if num_true_pos == 0:
return
a__ : Optional[int] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_exact.png" ) , title="Precision-Recall curve for Exact Match score" , )
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_f1.png" ) , title="Precision-Recall curve for F1 score" , )
a__ : str = {k: float(__a ) for k, v in qid_to_has_ans.items()}
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_oracle.png" ) , title="Oracle Precision-Recall curve (binary task of HasAns vs. NoAns)" , )
merge_eval(__a , __a , "pr_exact" )
merge_eval(__a , __a , "pr_f1" )
merge_eval(__a , __a , "pr_oracle" )
def UpperCamelCase_ ( __a , __a , __a , __a ) -> str:
if not qid_list:
return
a__ : Optional[Any] = [na_probs[k] for k in qid_list]
a__ : str = np.ones_like(__a ) / float(len(__a ) )
plt.hist(__a , weights=__a , bins=20 , range=(0.0, 1.0) )
plt.xlabel("Model probability of no-answer" )
plt.ylabel("Proportion of dataset" )
plt.title(f'''Histogram of no-answer probability: {name}''' )
plt.savefig(os.path.join(__a , f'''na_prob_hist_{name}.png''' ) )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[Any]:
a__ : str = sum(1 for k in qid_to_has_ans if not qid_to_has_ans[k] )
a__ : Optional[Any] = num_no_ans
a__ : Dict = cur_score
a__ : Any = 0.0
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
for i, qid in enumerate(__a ):
if qid not in scores:
continue
if qid_to_has_ans[qid]:
a__ : Optional[int] = scores[qid]
else:
if preds[qid]:
a__ : str = -1
else:
a__ : Union[str, Any] = 0
cur_score += diff
if cur_score > best_score:
a__ : Any = cur_score
a__ : Dict = na_probs[qid]
return 100.0 * best_score / len(__a ), best_thresh
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> Any:
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__ : Any = best_exact
a__ : Any = exact_thresh
a__ : List[Any] = best_fa
a__ : Optional[int] = fa_thresh
def UpperCamelCase_ ( ) -> Tuple:
with open(OPTS.data_file ) as f:
a__ : List[Any] = json.load(__a )
a__ : Any = dataset_json["data"]
with open(OPTS.pred_file ) as f:
a__ : int = json.load(__a )
if OPTS.na_prob_file:
with open(OPTS.na_prob_file ) as f:
a__ : List[str] = json.load(__a )
else:
a__ : Optional[int] = {k: 0.0 for k in preds}
a__ : Optional[Any] = make_qid_to_has_ans(__a ) # maps qid to True/False
a__ : List[Any] = [k for k, v in qid_to_has_ans.items() if v]
a__ : Union[str, Any] = [k for k, v in qid_to_has_ans.items() if not v]
a__, a__ : str = get_raw_scores(__a , __a )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : Tuple = make_eval_dict(__a , __a )
if has_ans_qids:
a__ : str = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "HasAns" )
if no_ans_qids:
a__ : List[Any] = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "NoAns" )
if OPTS.na_prob_file:
find_all_best_thresh(__a , __a , __a , __a , __a , __a )
if OPTS.na_prob_file and OPTS.out_image_dir:
run_precision_recall_analysis(__a , __a , __a , __a , __a , OPTS.out_image_dir )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "hasAns" )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "noAns" )
if OPTS.out_file:
with open(OPTS.out_file , "w" ) as f:
json.dump(__a , __a )
else:
print(json.dumps(__a , indent=2 ) )
if __name__ == "__main__":
UpperCamelCase : Any = parse_args()
if OPTS.out_image_dir:
import matplotlib
matplotlib.use("""Agg""")
import matplotlib.pyplot as plt
main()
| 37 |
from transformers import HfArgumentParser, TensorFlowBenchmark, TensorFlowBenchmarkArguments
def UpperCamelCase_ ( ) -> int:
a__ : Any = HfArgumentParser(__a )
a__ : Any = parser.parse_args_into_dataclasses()[0]
a__ : Optional[int] = TensorFlowBenchmark(args=__a )
try:
a__ : Optional[int] = parser.parse_args_into_dataclasses()[0]
except ValueError as e:
a__ : Tuple = "Arg --no_{0} is no longer used, please use --no-{0} instead."
a__ : List[Any] = " ".join(str(__a ).split(" " )[:-1] )
a__ : str = ""
a__ : List[Any] = eval(str(__a ).split(" " )[-1] )
a__ : List[str] = []
for arg in depreciated_args:
# arg[2:] removes '--'
if arg[2:] in TensorFlowBenchmark.deprecated_args:
# arg[5:] removes '--no_'
full_error_msg += arg_error_msg.format(arg[5:] )
else:
wrong_args.append(__a )
if len(__a ) > 0:
a__ : Tuple = full_error_msg + begin_error_msg + str(__a )
raise ValueError(__a )
benchmark.run()
if __name__ == "__main__":
main()
| 37 | 1 |
from . import (
albert,
align,
altclip,
audio_spectrogram_transformer,
auto,
autoformer,
bark,
bart,
barthez,
bartpho,
beit,
bert,
bert_generation,
bert_japanese,
bertweet,
big_bird,
bigbird_pegasus,
biogpt,
bit,
blenderbot,
blenderbot_small,
blip,
blip_a,
bloom,
bridgetower,
byta,
camembert,
canine,
chinese_clip,
clap,
clip,
clipseg,
codegen,
conditional_detr,
convbert,
convnext,
convnextva,
cpm,
cpmant,
ctrl,
cvt,
dataavec,
deberta,
deberta_va,
decision_transformer,
deformable_detr,
deit,
deprecated,
deta,
detr,
dialogpt,
dinat,
distilbert,
dit,
donut,
dpr,
dpt,
efficientformer,
efficientnet,
electra,
encodec,
encoder_decoder,
ernie,
ernie_m,
esm,
falcon,
flaubert,
flava,
fnet,
focalnet,
fsmt,
funnel,
git,
glpn,
gpta,
gpt_bigcode,
gpt_neo,
gpt_neox,
gpt_neox_japanese,
gpt_swa,
gptj,
gptsan_japanese,
graphormer,
groupvit,
herbert,
hubert,
ibert,
imagegpt,
informer,
instructblip,
jukebox,
layoutlm,
layoutlmva,
layoutlmva,
layoutxlm,
led,
levit,
lilt,
llama,
longformer,
longta,
luke,
lxmert,
mam_aaa,
marian,
markuplm,
maskaformer,
maskformer,
mbart,
mbartaa,
mega,
megatron_bert,
megatron_gpta,
mgp_str,
mluke,
mobilebert,
mobilenet_va,
mobilenet_va,
mobilevit,
mobilevitva,
mpnet,
mra,
mta,
musicgen,
mvp,
nat,
nezha,
nllb,
nllb_moe,
nystromformer,
oneformer,
open_llama,
openai,
opt,
owlvit,
pegasus,
pegasus_x,
perceiver,
phobert,
pixastruct,
plbart,
poolformer,
prophetnet,
qdqbert,
rag,
realm,
reformer,
regnet,
rembert,
resnet,
roberta,
roberta_prelayernorm,
roc_bert,
roformer,
rwkv,
sam,
segformer,
sew,
sew_d,
speech_encoder_decoder,
speech_to_text,
speech_to_text_a,
speechta,
splinter,
squeezebert,
swiftformer,
swin,
swinasr,
swinva,
switch_transformers,
ta,
table_transformer,
tapas,
time_series_transformer,
timesformer,
timm_backbone,
transfo_xl,
trocr,
tvlt,
umta,
unispeech,
unispeech_sat,
upernet,
videomae,
vilt,
vision_encoder_decoder,
vision_text_dual_encoder,
visual_bert,
vit,
vit_hybrid,
vit_mae,
vit_msn,
vivit,
wavaveca,
wavaveca_conformer,
wavaveca_phoneme,
wavaveca_with_lm,
wavlm,
whisper,
x_clip,
xglm,
xlm,
xlm_prophetnet,
xlm_roberta,
xlm_roberta_xl,
xlnet,
xmod,
yolos,
yoso,
)
| 37 |
import argparse
import ast
import logging
import os
import sys
import pandas as pd
import torch
from tqdm import tqdm
from transformers import BartForConditionalGeneration, RagRetriever, RagSequenceForGeneration, RagTokenForGeneration
from transformers import logging as transformers_logging
sys.path.append(os.path.join(os.getcwd())) # noqa: E402 # isort:skip
from utils_rag import exact_match_score, fa_score # noqa: E402 # isort:skip
UpperCamelCase : Optional[int] = logging.getLogger(__name__)
logging.basicConfig(level=logging.INFO)
transformers_logging.set_verbosity_info()
def UpperCamelCase_ ( __a ) -> Any:
if "token" in model_name_or_path:
return "rag_token"
if "sequence" in model_name_or_path:
return "rag_sequence"
if "bart" in model_name_or_path:
return "bart"
return None
def UpperCamelCase_ ( __a , __a , __a ) -> Any:
return max(metric_fn(__a , __a ) for gt in ground_truths )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = []
if args.gold_data_mode == "qa":
a__ : Any = pd.read_csv(__a , sep="\t" , header=__a )
for answer_list in data[1]:
a__ : Union[str, Any] = ast.literal_eval(__a )
answers.append(__a )
else:
a__ : List[str] = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : List[str] = [[reference] for reference in references]
a__ : List[str] = 0
for prediction, ground_truths in zip(__a , __a ):
total += 1
em += metric_max_over_ground_truths(__a , __a , __a )
fa += metric_max_over_ground_truths(__a , __a , __a )
a__ : Dict = 100.0 * em / total
a__ : Optional[Any] = 100.0 * fa / total
logger.info(f'''F1: {fa:.2f}''' )
logger.info(f'''EM: {em:.2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = args.k
a__ : str = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = 0
for hypo, reference in zip(__a , __a ):
a__ : Any = set(hypo.split("\t" )[:k] )
a__ : Union[str, Any] = set(reference.split("\t" ) )
total += 1
em += len(hypo_provenance & ref_provenance ) / k
a__ : Union[str, Any] = 100.0 * em / total
logger.info(f'''Precision@{k}: {em: .2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
def strip_title(__a ):
if title.startswith("\"" ):
a__ : Optional[Any] = title[1:]
if title.endswith("\"" ):
a__ : Union[str, Any] = title[:-1]
return title
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a , )["input_ids"].to(args.device )
a__ : Optional[int] = rag_model.rag.question_encoder(__a )
a__ : Union[str, Any] = question_enc_outputs[0]
a__ : Optional[int] = rag_model.retriever(
__a , question_enc_pool_output.cpu().detach().to(torch.floataa ).numpy() , prefix=rag_model.rag.generator.config.prefix , n_docs=rag_model.config.n_docs , return_tensors="pt" , )
a__ : List[Any] = rag_model.retriever.index.get_doc_dicts(result.doc_ids )
a__ : int = []
for docs in all_docs:
a__ : Optional[int] = [strip_title(__a ) for title in docs["title"]]
provenance_strings.append("\t".join(__a ) )
return provenance_strings
def UpperCamelCase_ ( __a , __a , __a ) -> Dict:
with torch.no_grad():
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a )
a__ : Any = inputs_dict.input_ids.to(args.device )
a__ : Dict = inputs_dict.attention_mask.to(args.device )
a__ : Optional[int] = rag_model.generate( # rag_model overwrites generate
__a , attention_mask=__a , num_beams=args.num_beams , min_length=args.min_length , max_length=args.max_length , early_stopping=__a , num_return_sequences=1 , bad_words_ids=[[0, 0]] , )
a__ : int = rag_model.retriever.generator_tokenizer.batch_decode(__a , skip_special_tokens=__a )
if args.print_predictions:
for q, a in zip(__a , __a ):
logger.info("Q: {} - A: {}".format(__a , __a ) )
return answers
def UpperCamelCase_ ( ) -> List[str]:
a__ : int = argparse.ArgumentParser()
parser.add_argument(
"--model_type" , choices=["rag_sequence", "rag_token", "bart"] , type=__a , help=(
"RAG model type: rag_sequence, rag_token or bart, if none specified, the type is inferred from the"
" model_name_or_path"
) , )
parser.add_argument(
"--index_name" , default=__a , choices=["exact", "compressed", "legacy"] , type=__a , help="RAG model retriever type" , )
parser.add_argument(
"--index_path" , default=__a , type=__a , help="Path to the retrieval index" , )
parser.add_argument("--n_docs" , default=5 , type=__a , help="Number of retrieved docs" )
parser.add_argument(
"--model_name_or_path" , default=__a , type=__a , required=__a , help="Path to pretrained checkpoints or model identifier from huggingface.co/models" , )
parser.add_argument(
"--eval_mode" , choices=["e2e", "retrieval"] , default="e2e" , type=__a , help=(
"Evaluation mode, e2e calculates exact match and F1 of the downstream task, retrieval calculates"
" precision@k."
) , )
parser.add_argument("--k" , default=1 , type=__a , help="k for the precision@k calculation" )
parser.add_argument(
"--evaluation_set" , default=__a , type=__a , required=__a , help="Path to a file containing evaluation samples" , )
parser.add_argument(
"--gold_data_path" , default=__a , type=__a , required=__a , help="Path to a tab-separated file with gold samples" , )
parser.add_argument(
"--gold_data_mode" , default="qa" , type=__a , choices=["qa", "ans"] , help=(
"Format of the gold data file"
"qa - a single line in the following format: question [tab] answer_list"
"ans - a single line of the gold file contains the expected answer string"
) , )
parser.add_argument(
"--predictions_path" , type=__a , default="predictions.txt" , help="Name of the predictions file, to be stored in the checkpoints directory" , )
parser.add_argument(
"--eval_all_checkpoints" , action="store_true" , help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number" , )
parser.add_argument(
"--eval_batch_size" , default=8 , type=__a , help="Batch size per GPU/CPU for evaluation." , )
parser.add_argument(
"--recalculate" , help="Recalculate predictions even if the prediction file exists" , action="store_true" , )
parser.add_argument(
"--num_beams" , default=4 , type=__a , help="Number of beams to be used when generating answers" , )
parser.add_argument("--min_length" , default=1 , type=__a , help="Min length of the generated answers" )
parser.add_argument("--max_length" , default=50 , type=__a , help="Max length of the generated answers" )
parser.add_argument(
"--print_predictions" , action="store_true" , help="If True, prints predictions while evaluating." , )
parser.add_argument(
"--print_docs" , action="store_true" , help="If True, prints docs retried while generating." , )
a__ : int = parser.parse_args()
a__ : Dict = torch.device("cuda" if torch.cuda.is_available() else "cpu" )
return args
def UpperCamelCase_ ( __a ) -> Optional[int]:
a__ : Tuple = {}
if args.model_type is None:
a__ : List[str] = infer_model_type(args.model_name_or_path )
assert args.model_type is not None
if args.model_type.startswith("rag" ):
a__ : int = RagTokenForGeneration if args.model_type == "rag_token" else RagSequenceForGeneration
a__ : Tuple = args.n_docs
if args.index_name is not None:
a__ : Any = args.index_name
if args.index_path is not None:
a__ : int = args.index_path
else:
a__ : Optional[Any] = BartForConditionalGeneration
a__ : Tuple = (
[f.path for f in os.scandir(args.model_name_or_path ) if f.is_dir()]
if args.eval_all_checkpoints
else [args.model_name_or_path]
)
logger.info("Evaluate the following checkpoints: %s" , __a )
a__ : Any = get_scores if args.eval_mode == "e2e" else get_precision_at_k
a__ : Union[str, Any] = evaluate_batch_eae if args.eval_mode == "e2e" else evaluate_batch_retrieval
for checkpoint in checkpoints:
if os.path.exists(args.predictions_path ) and (not args.recalculate):
logger.info("Calculating metrics based on an existing predictions file: {}".format(args.predictions_path ) )
score_fn(__a , args.predictions_path , args.gold_data_path )
continue
logger.info("***** Running evaluation for {} *****".format(__a ) )
logger.info(" Batch size = %d" , args.eval_batch_size )
logger.info(" Predictions will be stored under {}".format(args.predictions_path ) )
if args.model_type.startswith("rag" ):
a__ : str = RagRetriever.from_pretrained(__a , **__a )
a__ : Optional[int] = model_class.from_pretrained(__a , retriever=__a , **__a )
model.retriever.init_retrieval()
else:
a__ : Dict = model_class.from_pretrained(__a , **__a )
model.to(args.device )
with open(args.evaluation_set , "r" ) as eval_file, open(args.predictions_path , "w" ) as preds_file:
a__ : List[Any] = []
for line in tqdm(__a ):
questions.append(line.strip() )
if len(__a ) == args.eval_batch_size:
a__ : Union[str, Any] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) + "\n" )
preds_file.flush()
a__ : Any = []
if len(__a ) > 0:
a__ : List[str] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) )
preds_file.flush()
score_fn(__a , args.predictions_path , args.gold_data_path )
if __name__ == "__main__":
UpperCamelCase : List[Any] = get_args()
main(args)
| 37 | 1 |
import os
from typing import Dict, List, Union
import tensorflow as tf
from keras_nlp.tokenizers import BytePairTokenizer
from tensorflow_text import pad_model_inputs
from .tokenization_gpta import GPTaTokenizer
class A__ ( tf.keras.layers.Layer ):
"""simple docstring"""
def __init__( self : Dict , lowerCamelCase__ : Dict[str, int] , lowerCamelCase__ : List[str] , lowerCamelCase__ : int = None , lowerCamelCase__ : int = None ):
super().__init__()
a__ : Optional[Any] = pad_token_id
a__ : str = max_length
a__ : str = vocab
a__ : List[Any] = merges
a__ : List[Any] = BytePairTokenizer(lowerCamelCase__ , lowerCamelCase__ , sequence_length=lowerCamelCase__ )
@classmethod
def _UpperCamelCase( cls : Optional[int] , lowerCamelCase__ : GPTaTokenizer , *lowerCamelCase__ : Union[str, Any] , **lowerCamelCase__ : List[Any] ):
a__ : Optional[Any] = [" ".join(lowerCamelCase__ ) for m in tokenizer.bpe_ranks.keys()]
a__ : Optional[Any] = tokenizer.get_vocab()
return cls(lowerCamelCase__ , lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ )
@classmethod
def _UpperCamelCase( cls : Optional[Any] , lowerCamelCase__ : Union[str, os.PathLike] , *lowerCamelCase__ : Any , **lowerCamelCase__ : List[Any] ):
a__ : int = GPTaTokenizer.from_pretrained(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ )
return cls.from_tokenizer(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ )
@classmethod
def _UpperCamelCase( cls : Optional[int] , lowerCamelCase__ : int ):
return cls(**lowerCamelCase__ )
def _UpperCamelCase( self : List[str] ):
return {
"vocab": self.vocab,
"merges": self.merges,
"max_length": self.max_length,
"pad_token_id": self.pad_token_id,
}
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : int = None ):
a__ : Any = self.tf_tokenizer(lowerCamelCase__ )
a__ : Tuple = tf.ones_like(lowerCamelCase__ )
if self.pad_token_id is not None:
# pad the tokens up to max length
a__ : str = max_length if max_length is not None else self.max_length
if max_length is not None:
a__, a__ : Any = pad_model_inputs(
lowerCamelCase__ , max_seq_length=lowerCamelCase__ , pad_value=self.pad_token_id )
return {"attention_mask": attention_mask, "input_ids": input_ids}
| 37 |
import logging
import os
from dataclasses import dataclass, field
from typing import Dict, Optional
import datasets
import numpy as np
import tensorflow as tf
from transformers import (
AutoConfig,
AutoTokenizer,
EvalPrediction,
HfArgumentParser,
PreTrainedTokenizer,
TFAutoModelForSequenceClassification,
TFTrainer,
TFTrainingArguments,
)
from transformers.utils import logging as hf_logging
hf_logging.set_verbosity_info()
hf_logging.enable_default_handler()
hf_logging.enable_explicit_format()
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a = None , ) -> str:
a__ : int = {}
if train_file is not None:
a__ : int = [train_file]
if eval_file is not None:
a__ : Union[str, Any] = [eval_file]
if test_file is not None:
a__ : str = [test_file]
a__ : Optional[Any] = datasets.load_dataset("csv" , data_files=__a )
a__ : List[Any] = list(ds[list(files.keys() )[0]].features.keys() )
a__ : str = features_name.pop(__a )
a__ : Dict = list(set(ds[list(files.keys() )[0]][label_name] ) )
a__ : str = {label: i for i, label in enumerate(__a )}
a__ : Tuple = tokenizer.model_input_names
a__ : List[str] = {}
if len(__a ) == 1:
for k in files.keys():
a__ : Optional[Any] = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
example[features_name[0]] , truncation=__a , max_length=__a , padding="max_length" ) , batched=__a , )
elif len(__a ) == 2:
for k in files.keys():
a__ : Dict = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
(example[features_name[0]], example[features_name[1]]) , truncation=__a , max_length=__a , padding="max_length" , ) , batched=__a , )
def gen_train():
for ex in transformed_ds[datasets.Split.TRAIN]:
a__ : str = {k: v for k, v in ex.items() if k in input_names}
a__ : str = labelaid[ex[label_name]]
yield (d, label)
def gen_val():
for ex in transformed_ds[datasets.Split.VALIDATION]:
a__ : Tuple = {k: v for k, v in ex.items() if k in input_names}
a__ : List[Any] = labelaid[ex[label_name]]
yield (d, label)
def gen_test():
for ex in transformed_ds[datasets.Split.TEST]:
a__ : List[Any] = {k: v for k, v in ex.items() if k in input_names}
a__ : Optional[int] = labelaid[ex[label_name]]
yield (d, label)
a__ : Optional[Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TRAIN in transformed_ds
else None
)
if train_ds is not None:
a__ : Optional[int] = train_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TRAIN] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.VALIDATION in transformed_ds
else None
)
if val_ds is not None:
a__ : Optional[Any] = val_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.VALIDATION] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TEST in transformed_ds
else None
)
if test_ds is not None:
a__ : Tuple = test_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TEST] ) ) )
return train_ds, val_ds, test_ds, labelaid
UpperCamelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(metadata={'help': 'Which column contains the label'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the training file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the development file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the test file'} )
_lowercase = field(
default=1_2_8 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
_lowercase = field(
default=A__ , metadata={'help': 'Overwrite the cached training and evaluation sets'} )
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
_lowercase = field(default=A__ , metadata={'help': 'Set this flag to use fast tokenization.'} )
# If you want to tweak more attributes on your tokenizer, you should do it in a distinct script,
# or just modify its tokenizer_config.json.
_lowercase = field(
default=A__ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
def UpperCamelCase_ ( ) -> Union[str, Any]:
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
a__ : str = HfArgumentParser((ModelArguments, DataTrainingArguments, TFTrainingArguments) )
a__, a__, a__ : str = parser.parse_args_into_dataclasses()
if (
os.path.exists(training_args.output_dir )
and os.listdir(training_args.output_dir )
and training_args.do_train
and not training_args.overwrite_output_dir
):
raise ValueError(
f'''Output directory ({training_args.output_dir}) already exists and is not empty. Use'''
" --overwrite_output_dir to overcome." )
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s" , datefmt="%m/%d/%Y %H:%M:%S" , level=logging.INFO , )
logger.info(
f'''n_replicas: {training_args.n_replicas}, distributed training: {bool(training_args.n_replicas > 1 )}, '''
f'''16-bits training: {training_args.fpaa}''' )
logger.info(f'''Training/evaluation parameters {training_args}''' )
# Load pretrained model and tokenizer
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
a__ : Union[str, Any] = AutoTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , )
a__, a__, a__, a__ : Optional[Any] = get_tfds(
train_file=data_args.train_file , eval_file=data_args.dev_file , test_file=data_args.test_file , tokenizer=__a , label_column_id=data_args.label_column_id , max_seq_length=data_args.max_seq_length , )
a__ : Optional[int] = AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=len(__a ) , labelaid=__a , idalabel={id: label for label, id in labelaid.items()} , finetuning_task="text-classification" , cache_dir=model_args.cache_dir , )
with training_args.strategy.scope():
a__ : Any = TFAutoModelForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_pt=bool(".bin" in model_args.model_name_or_path ) , config=__a , cache_dir=model_args.cache_dir , )
def compute_metrics(__a ) -> Dict:
a__ : Union[str, Any] = np.argmax(p.predictions , axis=1 )
return {"acc": (preds == p.label_ids).mean()}
# Initialize our Trainer
a__ : Dict = TFTrainer(
model=__a , args=__a , train_dataset=__a , eval_dataset=__a , compute_metrics=__a , )
# Training
if training_args.do_train:
trainer.train()
trainer.save_model()
tokenizer.save_pretrained(training_args.output_dir )
# Evaluation
a__ : Optional[Any] = {}
if training_args.do_eval:
logger.info("*** Evaluate ***" )
a__ : Dict = trainer.evaluate()
a__ : int = os.path.join(training_args.output_dir , "eval_results.txt" )
with open(__a , "w" ) as writer:
logger.info("***** Eval results *****" )
for key, value in result.items():
logger.info(f''' {key} = {value}''' )
writer.write(f'''{key} = {value}\n''' )
results.update(__a )
return results
if __name__ == "__main__":
main()
| 37 | 1 |
import json
import os
import unittest
from transformers import CLIPTokenizer, CLIPTokenizerFast
from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_ftfy, require_tokenizers
from ...test_tokenization_common import TokenizerTesterMixin
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = CLIPTokenizer
_lowercase = CLIPTokenizerFast
_lowercase = True
_lowercase = {}
_lowercase = False
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# fmt: off
a__ : Any = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "lo", "l</w>", "w</w>", "r</w>", "t</w>", "low</w>", "er</w>", "lowest</w>", "newer</w>", "wider", "<unk>", "<|startoftext|>", "<|endoftext|>"]
# fmt: on
a__ : Optional[Any] = dict(zip(lowerCamelCase__ , range(len(lowerCamelCase__ ) ) ) )
a__ : Optional[Any] = ["#version: 0.2", "l o", "lo w</w>", "e r</w>"]
a__ : Optional[Any] = {"unk_token": "<unk>"}
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["vocab_file"] )
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["merges_file"] )
with open(self.vocab_file , "w" , encoding="utf-8" ) as fp:
fp.write(json.dumps(lowerCamelCase__ ) + "\n" )
with open(self.merges_file , "w" , encoding="utf-8" ) as fp:
fp.write("\n".join(lowerCamelCase__ ) )
def _UpperCamelCase( self : Dict , **lowerCamelCase__ : int ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizer.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , **lowerCamelCase__ : Optional[int] ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizerFast.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[Any] ):
a__ : int = "lower newer"
a__ : Optional[int] = "lower newer"
return input_text, output_text
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = CLIPTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
a__ : int = "lower newer"
a__ : List[str] = ["lo", "w", "er</w>", "n", "e", "w", "er</w>"]
a__ : Union[str, Any] = tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : int = tokens + [tokenizer.unk_token]
a__ : Union[str, Any] = [10, 2, 16, 9, 3, 2, 16, 20]
self.assertListEqual(tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , lowerCamelCase__ )
@require_ftfy
def _UpperCamelCase( self : Optional[Any] ):
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : List[str] = self.tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : Any = self.rust_tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : int = "A\n'll 11p223RF☆ho!!to?'d'd''d of a cat to-$''d."
a__ : Optional[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : Dict = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on an example containing a character (Latin Small Letter A
# with Tilde) encoded in 2 different ways
a__ : Optional[Any] = "xa\u0303y" + " " + "x\xe3y"
a__ : Optional[int] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of space type
a__ : str = [
"\u0009", # (horizontal tab, '\t')
"\u000B", # (vertical tab)
"\u000C", # (form feed)
"\u0020", # (space, ' ')
"\u200E", # (left-to-right mark):w
"\u200F", # (right-to-left mark)
]
for unicode_seq in spaces_unicodes:
a__ : Any = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of line break type
a__ : Union[str, Any] = [
"\u000A", # (line feed, '\n')
"\r\n", # (carriage return and line feed, '\r\n')
"\u000D", # (carriage return, '\r')
"\r", # (carriage return, '\r')
"\u000D", # (carriage return, '\r')
"\u2028", # (line separator)
"\u2029", # (paragraph separator)
# "\u0085", # (next line)
]
# The tokenization is not identical for the character "\u0085" (next line). The slow version using ftfy transforms
# it into the Horizontal Ellipsis character "…" ("\u2026") while the fast version transforms it into a
# space (and thus into an empty list).
for unicode_seq in line_break_unicodes:
a__ : List[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
# Test which aims to verify that the offsets are well adapted to the argument `add_prefix_space`
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : str = "hello" # `hello` is a token in the vocabulary of `pretrained_name`
a__ : Tuple = f'''{text_of_1_token} {text_of_1_token}'''
a__ : Optional[int] = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Union[str, Any] = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (0, len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (len(lowerCamelCase__ ) + 1, len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
a__ : Optional[Any] = f''' {text}'''
a__ : str = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Dict = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (1, 1 + len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (1 + len(lowerCamelCase__ ) + 1, 1 + len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
def _UpperCamelCase( self : int ):
# Test related to the breaking change introduced in transformers v4.17.0
# We need to check that an error in raised when the user try to load a previous version of the tokenizer.
with self.assertRaises(lowerCamelCase__ ) as context:
self.rust_tokenizer_class.from_pretrained("robot-test/old-clip-tokenizer" )
self.assertTrue(
context.exception.args[0].startswith(
"The `backend_tokenizer` provided does not match the expected format." ) )
@require_ftfy
def _UpperCamelCase( self : int ):
super().test_tokenization_python_rust_equals()
def _UpperCamelCase( self : str ):
# CLIP always lower cases letters
pass
| 37 |
import argparse
import collections
import json
import os
import re
import string
import sys
import numpy as np
UpperCamelCase : List[str] = re.compile(r"""\b(a|an|the)\b""", re.UNICODE)
UpperCamelCase : Union[str, Any] = None
def UpperCamelCase_ ( ) -> List[str]:
a__ : List[Any] = argparse.ArgumentParser("Official evaluation script for SQuAD version 2.0." )
parser.add_argument("data_file" , metavar="data.json" , help="Input data JSON file." )
parser.add_argument("pred_file" , metavar="pred.json" , help="Model predictions." )
parser.add_argument(
"--out-file" , "-o" , metavar="eval.json" , help="Write accuracy metrics to file (default is stdout)." )
parser.add_argument(
"--na-prob-file" , "-n" , metavar="na_prob.json" , help="Model estimates of probability of no answer." )
parser.add_argument(
"--na-prob-thresh" , "-t" , type=__a , default=1.0 , help="Predict \"\" if no-answer probability exceeds this (default = 1.0)." , )
parser.add_argument(
"--out-image-dir" , "-p" , metavar="out_images" , default=__a , help="Save precision-recall curves to directory." )
parser.add_argument("--verbose" , "-v" , action="store_true" )
if len(sys.argv ) == 1:
parser.print_help()
sys.exit(1 )
return parser.parse_args()
def UpperCamelCase_ ( __a ) -> str:
a__ : Optional[Any] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : Dict = bool(qa["answers"]["text"] )
return qid_to_has_ans
def UpperCamelCase_ ( __a ) -> List[Any]:
def remove_articles(__a ):
return ARTICLES_REGEX.sub(" " , __a )
def white_space_fix(__a ):
return " ".join(text.split() )
def remove_punc(__a ):
a__ : Union[str, Any] = set(string.punctuation )
return "".join(ch for ch in text if ch not in exclude )
def lower(__a ):
return text.lower()
return white_space_fix(remove_articles(remove_punc(lower(__a ) ) ) )
def UpperCamelCase_ ( __a ) -> Dict:
if not s:
return []
return normalize_answer(__a ).split()
def UpperCamelCase_ ( __a , __a ) -> str:
return int(normalize_answer(__a ) == normalize_answer(__a ) )
def UpperCamelCase_ ( __a , __a ) -> Dict:
a__ : int = get_tokens(__a )
a__ : Optional[Any] = get_tokens(__a )
a__ : Any = collections.Counter(__a ) & collections.Counter(__a )
a__ : Dict = sum(common.values() )
if len(__a ) == 0 or len(__a ) == 0:
# If either is no-answer, then F1 is 1 if they agree, 0 otherwise
return int(gold_toks == pred_toks )
if num_same == 0:
return 0
a__ : Tuple = 1.0 * num_same / len(__a )
a__ : str = 1.0 * num_same / len(__a )
a__ : str = (2 * precision * recall) / (precision + recall)
return fa
def UpperCamelCase_ ( __a , __a ) -> int:
a__ : List[str] = {}
a__ : Optional[int] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : List[Any] = qa["id"]
a__ : Dict = [t for t in qa["answers"]["text"] if normalize_answer(__a )]
if not gold_answers:
# For unanswerable questions, only correct answer is empty string
a__ : Tuple = [""]
if qid not in preds:
print(f'''Missing prediction for {qid}''' )
continue
a__ : Tuple = preds[qid]
# Take max over all gold answers
a__ : Optional[int] = max(compute_exact(__a , __a ) for a in gold_answers )
a__ : str = max(compute_fa(__a , __a ) for a in gold_answers )
return exact_scores, fa_scores
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
a__ : Optional[Any] = {}
for qid, s in scores.items():
a__ : Dict = na_probs[qid] > na_prob_thresh
if pred_na:
a__ : Dict = float(not qid_to_has_ans[qid] )
else:
a__ : Optional[Any] = s
return new_scores
def UpperCamelCase_ ( __a , __a , __a=None ) -> Tuple:
if not qid_list:
a__ : Union[str, Any] = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores.values() ) / total),
("f1", 100.0 * sum(fa_scores.values() ) / total),
("total", total),
] )
else:
a__ : int = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores[k] for k in qid_list ) / total),
("f1", 100.0 * sum(fa_scores[k] for k in qid_list ) / total),
("total", total),
] )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
for k in new_eval:
a__ : Optional[Any] = new_eval[k]
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
plt.step(__a , __a , color="b" , alpha=0.2 , where="post" )
plt.fill_between(__a , __a , step="post" , alpha=0.2 , color="b" )
plt.xlabel("Recall" )
plt.ylabel("Precision" )
plt.xlim([0.0, 1.05] )
plt.ylim([0.0, 1.05] )
plt.title(__a )
plt.savefig(__a )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a , __a=None , __a=None ) -> Dict:
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
a__ : Any = 0.0
a__ : Optional[int] = 1.0
a__ : Optional[int] = 0.0
a__ : Any = [1.0]
a__ : Tuple = [0.0]
a__ : List[str] = 0.0
for i, qid in enumerate(__a ):
if qid_to_has_ans[qid]:
true_pos += scores[qid]
a__ : Any = true_pos / float(i + 1 )
a__ : int = true_pos / float(__a )
if i == len(__a ) - 1 or na_probs[qid] != na_probs[qid_list[i + 1]]:
# i.e., if we can put a threshold after this point
avg_prec += cur_p * (cur_r - recalls[-1])
precisions.append(__a )
recalls.append(__a )
if out_image:
plot_pr_curve(__a , __a , __a , __a )
return {"ap": 100.0 * avg_prec}
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> str:
if out_image_dir and not os.path.exists(__a ):
os.makedirs(__a )
a__ : Optional[int] = sum(1 for v in qid_to_has_ans.values() if v )
if num_true_pos == 0:
return
a__ : Optional[int] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_exact.png" ) , title="Precision-Recall curve for Exact Match score" , )
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_f1.png" ) , title="Precision-Recall curve for F1 score" , )
a__ : str = {k: float(__a ) for k, v in qid_to_has_ans.items()}
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_oracle.png" ) , title="Oracle Precision-Recall curve (binary task of HasAns vs. NoAns)" , )
merge_eval(__a , __a , "pr_exact" )
merge_eval(__a , __a , "pr_f1" )
merge_eval(__a , __a , "pr_oracle" )
def UpperCamelCase_ ( __a , __a , __a , __a ) -> str:
if not qid_list:
return
a__ : Optional[Any] = [na_probs[k] for k in qid_list]
a__ : str = np.ones_like(__a ) / float(len(__a ) )
plt.hist(__a , weights=__a , bins=20 , range=(0.0, 1.0) )
plt.xlabel("Model probability of no-answer" )
plt.ylabel("Proportion of dataset" )
plt.title(f'''Histogram of no-answer probability: {name}''' )
plt.savefig(os.path.join(__a , f'''na_prob_hist_{name}.png''' ) )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[Any]:
a__ : str = sum(1 for k in qid_to_has_ans if not qid_to_has_ans[k] )
a__ : Optional[Any] = num_no_ans
a__ : Dict = cur_score
a__ : Any = 0.0
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
for i, qid in enumerate(__a ):
if qid not in scores:
continue
if qid_to_has_ans[qid]:
a__ : Optional[int] = scores[qid]
else:
if preds[qid]:
a__ : str = -1
else:
a__ : Union[str, Any] = 0
cur_score += diff
if cur_score > best_score:
a__ : Any = cur_score
a__ : Dict = na_probs[qid]
return 100.0 * best_score / len(__a ), best_thresh
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> Any:
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__ : Any = best_exact
a__ : Any = exact_thresh
a__ : List[Any] = best_fa
a__ : Optional[int] = fa_thresh
def UpperCamelCase_ ( ) -> Tuple:
with open(OPTS.data_file ) as f:
a__ : List[Any] = json.load(__a )
a__ : Any = dataset_json["data"]
with open(OPTS.pred_file ) as f:
a__ : int = json.load(__a )
if OPTS.na_prob_file:
with open(OPTS.na_prob_file ) as f:
a__ : List[str] = json.load(__a )
else:
a__ : Optional[int] = {k: 0.0 for k in preds}
a__ : Optional[Any] = make_qid_to_has_ans(__a ) # maps qid to True/False
a__ : List[Any] = [k for k, v in qid_to_has_ans.items() if v]
a__ : Union[str, Any] = [k for k, v in qid_to_has_ans.items() if not v]
a__, a__ : str = get_raw_scores(__a , __a )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : Tuple = make_eval_dict(__a , __a )
if has_ans_qids:
a__ : str = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "HasAns" )
if no_ans_qids:
a__ : List[Any] = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "NoAns" )
if OPTS.na_prob_file:
find_all_best_thresh(__a , __a , __a , __a , __a , __a )
if OPTS.na_prob_file and OPTS.out_image_dir:
run_precision_recall_analysis(__a , __a , __a , __a , __a , OPTS.out_image_dir )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "hasAns" )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "noAns" )
if OPTS.out_file:
with open(OPTS.out_file , "w" ) as f:
json.dump(__a , __a )
else:
print(json.dumps(__a , indent=2 ) )
if __name__ == "__main__":
UpperCamelCase : Any = parse_args()
if OPTS.out_image_dir:
import matplotlib
matplotlib.use("""Agg""")
import matplotlib.pyplot as plt
main()
| 37 | 1 |
import inspect
import unittest
from huggingface_hub import hf_hub_download
from transformers import ASTConfig
from transformers.testing_utils import require_torch, require_torchaudio, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_torchaudio_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import ASTForAudioClassification, ASTModel
from transformers.models.audio_spectrogram_transformer.modeling_audio_spectrogram_transformer import (
AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
)
if is_torchaudio_available():
import torchaudio
from transformers import ASTFeatureExtractor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[str] , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[Any]=2 , lowerCamelCase__ : Any=24 , lowerCamelCase__ : Optional[Any]=16 , lowerCamelCase__ : int=True , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[Any]=32 , lowerCamelCase__ : List[str]=5 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Optional[Any]=37 , lowerCamelCase__ : Any="gelu" , lowerCamelCase__ : Union[str, Any]=0.1 , lowerCamelCase__ : Optional[int]=0.1 , lowerCamelCase__ : str=10 , lowerCamelCase__ : Optional[Any]=0.02 , lowerCamelCase__ : str=None , lowerCamelCase__ : List[str]=2 , lowerCamelCase__ : Optional[Any]=2 , ):
a__ : str = parent
a__ : Any = batch_size
a__ : Dict = patch_size
a__ : List[Any] = max_length
a__ : str = num_mel_bins
a__ : Optional[Any] = is_training
a__ : Optional[int] = use_labels
a__ : List[Any] = hidden_size
a__ : str = num_hidden_layers
a__ : Any = num_attention_heads
a__ : Union[str, Any] = intermediate_size
a__ : List[str] = hidden_act
a__ : str = hidden_dropout_prob
a__ : Tuple = attention_probs_dropout_prob
a__ : List[Any] = type_sequence_label_size
a__ : Any = initializer_range
a__ : str = scope
a__ : List[str] = frequency_stride
a__ : Union[str, Any] = time_stride
# in AST, the seq length equals the number of patches + 2 (we add 2 for the [CLS] and distillation tokens)
a__ : List[Any] = (self.num_mel_bins - self.patch_size) // self.frequency_stride + 1
a__ : List[str] = (self.max_length - self.patch_size) // self.time_stride + 1
a__ : Tuple = frequency_out_dimension * time_out_dimension
a__ : List[str] = num_patches + 2
def _UpperCamelCase( self : List[str] ):
a__ : Any = floats_tensor([self.batch_size, self.max_length, self.num_mel_bins] )
a__ : List[Any] = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : List[str] = self.get_config()
return config, input_values, labels
def _UpperCamelCase( self : Optional[int] ):
return ASTConfig(
patch_size=self.patch_size , max_length=self.max_length , num_mel_bins=self.num_mel_bins , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , frequency_stride=self.frequency_stride , time_stride=self.time_stride , )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : int , lowerCamelCase__ : Optional[int] ):
a__ : List[Any] = ASTModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Dict = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : str ):
a__ : Dict = self.prepare_config_and_inputs()
(
(
a__
), (
a__
), (
a__
),
) : Optional[int] = config_and_inputs
a__ : List[Any] = {"input_values": input_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(
ASTModel,
ASTForAudioClassification,
)
if is_torch_available()
else ()
)
_lowercase = (
{'audio-classification': ASTForAudioClassification, 'feature-extraction': ASTModel}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
if pipeline_test_casse_name == "AudioClassificationPipelineTests":
return True
return False
def _UpperCamelCase( self : str ):
a__ : str = ASTModelTester(self )
a__ : Any = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[str] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="AST does not use inputs_embeds" )
def _UpperCamelCase( self : List[str] ):
pass
def _UpperCamelCase( self : Optional[int] ):
a__, a__ : str = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Any = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Union[str, Any] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : Tuple ):
a__, a__ : List[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Dict = model_class(lowerCamelCase__ )
a__ : Optional[int] = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Optional[Any] = ["input_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
for model_name in AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Union[str, Any] = ASTModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : Optional[int] = hf_hub_download(
repo_id="nielsr/audio-spectogram-transformer-checkpoint" , filename="sample_audio.flac" , repo_type="dataset" )
a__, a__ : List[str] = torchaudio.load(__a )
return audio, sampling_rate
@require_torch
@require_torchaudio
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : List[str] ):
return (
ASTFeatureExtractor.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" )
if is_torchaudio_available()
else None
)
@slow
def _UpperCamelCase( self : Optional[int] ):
a__ : int = self.default_feature_extractor
a__ : Optional[Any] = ASTForAudioClassification.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" ).to(lowerCamelCase__ )
a__ : Any = self.default_feature_extractor
a__, a__ : Dict = prepare_audio()
a__ : str = audio.squeeze().numpy()
a__ : Any = feature_extractor(lowerCamelCase__ , sampling_rate=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(**lowerCamelCase__ )
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 527) )
self.assertEqual(outputs.logits.shape , lowerCamelCase__ )
a__ : List[str] = torch.tensor([-0.8760, -7.0042, -8.6602] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(outputs.logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
| 37 |
import json
import os
import unittest
from transformers import CLIPTokenizer, CLIPTokenizerFast
from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_ftfy, require_tokenizers
from ...test_tokenization_common import TokenizerTesterMixin
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = CLIPTokenizer
_lowercase = CLIPTokenizerFast
_lowercase = True
_lowercase = {}
_lowercase = False
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# fmt: off
a__ : Any = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "lo", "l</w>", "w</w>", "r</w>", "t</w>", "low</w>", "er</w>", "lowest</w>", "newer</w>", "wider", "<unk>", "<|startoftext|>", "<|endoftext|>"]
# fmt: on
a__ : Optional[Any] = dict(zip(lowerCamelCase__ , range(len(lowerCamelCase__ ) ) ) )
a__ : Optional[Any] = ["#version: 0.2", "l o", "lo w</w>", "e r</w>"]
a__ : Optional[Any] = {"unk_token": "<unk>"}
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["vocab_file"] )
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["merges_file"] )
with open(self.vocab_file , "w" , encoding="utf-8" ) as fp:
fp.write(json.dumps(lowerCamelCase__ ) + "\n" )
with open(self.merges_file , "w" , encoding="utf-8" ) as fp:
fp.write("\n".join(lowerCamelCase__ ) )
def _UpperCamelCase( self : Dict , **lowerCamelCase__ : int ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizer.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , **lowerCamelCase__ : Optional[int] ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizerFast.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[Any] ):
a__ : int = "lower newer"
a__ : Optional[int] = "lower newer"
return input_text, output_text
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = CLIPTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
a__ : int = "lower newer"
a__ : List[str] = ["lo", "w", "er</w>", "n", "e", "w", "er</w>"]
a__ : Union[str, Any] = tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : int = tokens + [tokenizer.unk_token]
a__ : Union[str, Any] = [10, 2, 16, 9, 3, 2, 16, 20]
self.assertListEqual(tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , lowerCamelCase__ )
@require_ftfy
def _UpperCamelCase( self : Optional[Any] ):
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : List[str] = self.tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : Any = self.rust_tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : int = "A\n'll 11p223RF☆ho!!to?'d'd''d of a cat to-$''d."
a__ : Optional[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : Dict = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on an example containing a character (Latin Small Letter A
# with Tilde) encoded in 2 different ways
a__ : Optional[Any] = "xa\u0303y" + " " + "x\xe3y"
a__ : Optional[int] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of space type
a__ : str = [
"\u0009", # (horizontal tab, '\t')
"\u000B", # (vertical tab)
"\u000C", # (form feed)
"\u0020", # (space, ' ')
"\u200E", # (left-to-right mark):w
"\u200F", # (right-to-left mark)
]
for unicode_seq in spaces_unicodes:
a__ : Any = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of line break type
a__ : Union[str, Any] = [
"\u000A", # (line feed, '\n')
"\r\n", # (carriage return and line feed, '\r\n')
"\u000D", # (carriage return, '\r')
"\r", # (carriage return, '\r')
"\u000D", # (carriage return, '\r')
"\u2028", # (line separator)
"\u2029", # (paragraph separator)
# "\u0085", # (next line)
]
# The tokenization is not identical for the character "\u0085" (next line). The slow version using ftfy transforms
# it into the Horizontal Ellipsis character "…" ("\u2026") while the fast version transforms it into a
# space (and thus into an empty list).
for unicode_seq in line_break_unicodes:
a__ : List[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
# Test which aims to verify that the offsets are well adapted to the argument `add_prefix_space`
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : str = "hello" # `hello` is a token in the vocabulary of `pretrained_name`
a__ : Tuple = f'''{text_of_1_token} {text_of_1_token}'''
a__ : Optional[int] = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Union[str, Any] = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (0, len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (len(lowerCamelCase__ ) + 1, len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
a__ : Optional[Any] = f''' {text}'''
a__ : str = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Dict = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (1, 1 + len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (1 + len(lowerCamelCase__ ) + 1, 1 + len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
def _UpperCamelCase( self : int ):
# Test related to the breaking change introduced in transformers v4.17.0
# We need to check that an error in raised when the user try to load a previous version of the tokenizer.
with self.assertRaises(lowerCamelCase__ ) as context:
self.rust_tokenizer_class.from_pretrained("robot-test/old-clip-tokenizer" )
self.assertTrue(
context.exception.args[0].startswith(
"The `backend_tokenizer` provided does not match the expected format." ) )
@require_ftfy
def _UpperCamelCase( self : int ):
super().test_tokenization_python_rust_equals()
def _UpperCamelCase( self : str ):
# CLIP always lower cases letters
pass
| 37 | 1 |
import gc
import tempfile
import unittest
import numpy as np
import torch
from diffusers import VersatileDiffusionTextToImagePipeline
from diffusers.utils.testing_utils import nightly, require_torch_gpu, torch_device
UpperCamelCase : Any = False
class A__ ( unittest.TestCase ):
"""simple docstring"""
pass
@nightly
@require_torch_gpu
class A__ ( unittest.TestCase ):
"""simple docstring"""
def _UpperCamelCase( self : Optional[Any] ):
# clean up the VRAM after each test
super().tearDown()
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : str ):
a__ : Tuple = VersatileDiffusionTextToImagePipeline.from_pretrained("shi-labs/versatile-diffusion" )
# remove text_unet
pipe.remove_unused_weights()
pipe.to(lowerCamelCase__ )
pipe.set_progress_bar_config(disable=lowerCamelCase__ )
a__ : Tuple = "A painting of a squirrel eating a burger "
a__ : Any = torch.manual_seed(0 )
a__ : int = pipe(
prompt=lowerCamelCase__ , generator=lowerCamelCase__ , guidance_scale=7.5 , num_inference_steps=2 , output_type="numpy" ).images
with tempfile.TemporaryDirectory() as tmpdirname:
pipe.save_pretrained(lowerCamelCase__ )
a__ : int = VersatileDiffusionTextToImagePipeline.from_pretrained(lowerCamelCase__ )
pipe.to(lowerCamelCase__ )
pipe.set_progress_bar_config(disable=lowerCamelCase__ )
a__ : Optional[int] = generator.manual_seed(0 )
a__ : Optional[int] = pipe(
prompt=lowerCamelCase__ , generator=lowerCamelCase__ , guidance_scale=7.5 , num_inference_steps=2 , output_type="numpy" ).images
assert np.abs(image - new_image ).sum() < 1E-5, "Models don't have the same forward pass"
def _UpperCamelCase( self : Tuple ):
a__ : List[str] = VersatileDiffusionTextToImagePipeline.from_pretrained(
"shi-labs/versatile-diffusion" , torch_dtype=torch.floataa )
pipe.to(lowerCamelCase__ )
pipe.set_progress_bar_config(disable=lowerCamelCase__ )
a__ : Optional[Any] = "A painting of a squirrel eating a burger "
a__ : Dict = torch.manual_seed(0 )
a__ : int = pipe(
prompt=lowerCamelCase__ , generator=lowerCamelCase__ , guidance_scale=7.5 , num_inference_steps=50 , output_type="numpy" ).images
a__ : Union[str, Any] = image[0, 253:256, 253:256, -1]
assert image.shape == (1, 512, 512, 3)
a__ : List[Any] = np.array([0.3367, 0.3169, 0.2656, 0.3870, 0.4790, 0.3796, 0.4009, 0.4878, 0.4778] )
assert np.abs(image_slice.flatten() - expected_slice ).max() < 1E-2
| 37 |
import os
import re
import shutil
from argparse import ArgumentParser, Namespace
from datasets.commands import BaseDatasetsCLICommand
from datasets.utils.logging import get_logger
UpperCamelCase : Dict = """<<<<<<< This should probably be modified because it mentions: """
UpperCamelCase : List[Any] = """=======
>>>>>>>
"""
UpperCamelCase : Optional[Any] = [
"""TextEncoderConfig""",
"""ByteTextEncoder""",
"""SubwordTextEncoder""",
"""encoder_config""",
"""maybe_build_from_corpus""",
"""manual_dir""",
]
UpperCamelCase : Any = [
# (pattern, replacement)
# Order is important here for some replacements
(r"""tfds\.core""", r"""datasets"""),
(r"""tf\.io\.gfile\.GFile""", r"""open"""),
(r"""tf\.([\w\d]+)""", r"""datasets.Value('\1')"""),
(r"""tfds\.features\.Text\(\)""", r"""datasets.Value('string')"""),
(r"""tfds\.features\.Text\(""", r"""datasets.Value('string'),"""),
(r"""features\s*=\s*tfds.features.FeaturesDict\(""", r"""features=datasets.Features("""),
(r"""tfds\.features\.FeaturesDict\(""", r"""dict("""),
(r"""The TensorFlow Datasets Authors""", r"""The TensorFlow Datasets Authors and the HuggingFace Datasets Authors"""),
(r"""tfds\.""", r"""datasets."""),
(r"""dl_manager\.manual_dir""", r"""self.config.data_dir"""),
(r"""self\.builder_config""", r"""self.config"""),
]
def UpperCamelCase_ ( __a ) -> Optional[Any]:
return ConvertCommand(args.tfds_path , args.datasets_directory )
class A__ ( A__ ):
"""simple docstring"""
@staticmethod
def _UpperCamelCase( lowerCamelCase__ : ArgumentParser ):
a__ : List[str] = parser.add_parser(
"convert" , help="Convert a TensorFlow Datasets dataset to a HuggingFace Datasets dataset." , )
train_parser.add_argument(
"--tfds_path" , type=lowerCamelCase__ , required=lowerCamelCase__ , help="Path to a TensorFlow Datasets folder to convert or a single tfds file to convert." , )
train_parser.add_argument(
"--datasets_directory" , type=lowerCamelCase__ , required=lowerCamelCase__ , help="Path to the HuggingFace Datasets folder." )
train_parser.set_defaults(func=lowerCamelCase__ )
def __init__( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str , *lowerCamelCase__ : Tuple ):
a__ : str = get_logger("datasets-cli/converting" )
a__ : Optional[Any] = tfds_path
a__ : Optional[int] = datasets_directory
def _UpperCamelCase( self : int ):
if os.path.isdir(self._tfds_path ):
a__ : List[str] = os.path.abspath(self._tfds_path )
elif os.path.isfile(self._tfds_path ):
a__ : Any = os.path.dirname(self._tfds_path )
else:
raise ValueError("--tfds_path is neither a directory nor a file. Please check path." )
a__ : Dict = os.path.abspath(self._datasets_directory )
self._logger.info(f'''Converting datasets from {abs_tfds_path} to {abs_datasets_path}''' )
a__ : Tuple = []
a__ : str = []
a__ : List[Any] = {}
if os.path.isdir(self._tfds_path ):
a__ : List[str] = os.listdir(lowerCamelCase__ )
else:
a__ : Union[str, Any] = [os.path.basename(self._tfds_path )]
for f_name in file_names:
self._logger.info(f'''Looking at file {f_name}''' )
a__ : Any = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
a__ : Dict = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
if not os.path.isfile(lowerCamelCase__ ) or "__init__" in f_name or "_test" in f_name or ".py" not in f_name:
self._logger.info("Skipping file" )
continue
with open(lowerCamelCase__ , encoding="utf-8" ) as f:
a__ : List[Any] = f.readlines()
a__ : Union[str, Any] = []
a__ : Union[str, Any] = False
a__ : Union[str, Any] = False
a__ : Dict = []
for line in lines:
a__ : Optional[Any] = line
# Convert imports
if "import tensorflow.compat.v2 as tf" in out_line:
continue
elif "@tfds.core" in out_line:
continue
elif "builder=self" in out_line:
continue
elif "import tensorflow_datasets.public_api as tfds" in out_line:
a__ : List[Any] = "import datasets\n"
elif "import tensorflow" in out_line:
# order is important here
a__ : List[str] = ""
continue
elif "from absl import logging" in out_line:
a__ : Dict = "from datasets import logging\n"
elif "getLogger" in out_line:
a__ : List[Any] = out_line.replace("getLogger" , "get_logger" )
elif any(expression in out_line for expression in TO_HIGHLIGHT ):
a__ : List[str] = True
a__ : Dict = list(filter(lambda lowerCamelCase__ : e in out_line , lowerCamelCase__ ) )
out_lines.append(HIGHLIGHT_MESSAGE_PRE + str(lowerCamelCase__ ) + "\n" )
out_lines.append(lowerCamelCase__ )
out_lines.append(lowerCamelCase__ )
continue
else:
for pattern, replacement in TO_CONVERT:
a__ : Tuple = re.sub(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
# Take care of saving utilities (to later move them together with main script)
if "tensorflow_datasets" in out_line:
a__ : Optional[int] = re.match(r"from\stensorflow_datasets.*import\s([^\.\r\n]+)" , lowerCamelCase__ )
tfds_imports.extend(imp.strip() for imp in match.group(1 ).split("," ) )
a__ : Optional[Any] = "from . import " + match.group(1 )
# Check we have not forget anything
if "tf." in out_line or "tfds." in out_line or "tensorflow_datasets" in out_line:
raise ValueError(f'''Error converting {out_line.strip()}''' )
if "GeneratorBasedBuilder" in out_line or "BeamBasedBuilder" in out_line:
a__ : Optional[int] = True
out_lines.append(lowerCamelCase__ )
if is_builder or "wmt" in f_name:
# We create a new directory for each dataset
a__ : Dict = f_name.replace(".py" , "" )
a__ : Optional[int] = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
a__ : Any = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
os.makedirs(lowerCamelCase__ , exist_ok=lowerCamelCase__ )
self._logger.info(f'''Adding directory {output_dir}''' )
imports_to_builder_map.update({imp: output_dir for imp in tfds_imports} )
else:
# Utilities will be moved at the end
utils_files.append(lowerCamelCase__ )
if needs_manual_update:
with_manual_update.append(lowerCamelCase__ )
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as f:
f.writelines(lowerCamelCase__ )
self._logger.info(f'''Converted in {output_file}''' )
for utils_file in utils_files:
try:
a__ : Any = os.path.basename(lowerCamelCase__ )
a__ : Optional[int] = imports_to_builder_map[f_name.replace(".py" , "" )]
self._logger.info(f'''Moving {dest_folder} to {utils_file}''' )
shutil.copy(lowerCamelCase__ , lowerCamelCase__ )
except KeyError:
self._logger.error(f'''Cannot find destination folder for {utils_file}. Please copy manually.''' )
if with_manual_update:
for file_path in with_manual_update:
self._logger.warning(
f'''You need to manually update file {file_path} to remove configurations using \'TextEncoderConfig\'.''' )
| 37 | 1 |
from typing import Dict, List, Optional, Union
import numpy as np
from ...image_processing_utils import BaseImageProcessor, BatchFeature, get_size_dict
from ...image_transforms import (
center_crop,
get_resize_output_image_size,
normalize,
rescale,
resize,
to_channel_dimension_format,
)
from ...image_utils import (
IMAGENET_DEFAULT_MEAN,
IMAGENET_DEFAULT_STD,
ChannelDimension,
ImageInput,
PILImageResampling,
make_list_of_images,
to_numpy_array,
valid_images,
)
from ...utils import TensorType, is_vision_available, logging
if is_vision_available():
import PIL
UpperCamelCase : int = logging.get_logger(__name__)
class A__ ( A__ ):
"""simple docstring"""
_lowercase = ['pixel_values']
def __init__( self : int , lowerCamelCase__ : bool = True , lowerCamelCase__ : Dict[str, int] = None , lowerCamelCase__ : int = 0.9 , lowerCamelCase__ : PILImageResampling = PILImageResampling.BICUBIC , lowerCamelCase__ : bool = True , lowerCamelCase__ : Dict[str, int] = None , lowerCamelCase__ : Union[int, float] = 1 / 255 , lowerCamelCase__ : bool = True , lowerCamelCase__ : bool = True , lowerCamelCase__ : Optional[Union[float, List[float]]] = None , lowerCamelCase__ : Optional[Union[float, List[float]]] = None , **lowerCamelCase__ : Any , ):
super().__init__(**lowerCamelCase__ )
a__ : str = size if size is not None else {"shortest_edge": 224}
a__ : str = get_size_dict(lowerCamelCase__ , default_to_square=lowerCamelCase__ )
a__ : Dict = crop_size if crop_size is not None else {"height": 224, "width": 224}
a__ : List[Any] = get_size_dict(lowerCamelCase__ , param_name="crop_size" )
a__ : Optional[Any] = do_resize
a__ : str = size
a__ : List[Any] = crop_pct
a__ : Tuple = resample
a__ : int = do_center_crop
a__ : Union[str, Any] = crop_size
a__ : Any = do_rescale
a__ : Optional[int] = rescale_factor
a__ : Optional[Any] = do_normalize
a__ : Dict = image_mean if image_mean is not None else IMAGENET_DEFAULT_MEAN
a__ : Dict = image_std if image_std is not None else IMAGENET_DEFAULT_STD
def _UpperCamelCase( self : Dict , lowerCamelCase__ : np.ndarray , lowerCamelCase__ : Dict[str, int] , lowerCamelCase__ : Optional[float] = None , lowerCamelCase__ : PILImageResampling = PILImageResampling.BICUBIC , lowerCamelCase__ : Optional[Union[str, ChannelDimension]] = None , **lowerCamelCase__ : str , ):
a__ : Optional[int] = get_size_dict(lowerCamelCase__ , default_to_square=lowerCamelCase__ )
if "shortest_edge" not in size and ("height" not in size or "width" not in size):
raise ValueError(f'''size must contain \'height\' and \'width\' or \'shortest_edge\' as keys. Got {size.keys()}''' )
if crop_pct is not None:
if "shortest_edge" in size:
a__ : Dict = int(size["shortest_edge"] / crop_pct )
elif "height" in size and "width" in size:
if size["height"] == size["width"]:
a__ : Optional[Any] = int(size["height"] / crop_pct )
else:
a__ : List[str] = (int(size["height"] / crop_pct ), int(size["width"] / crop_pct ))
else:
raise ValueError("Invalid size for resize: {}".format(lowerCamelCase__ ) )
a__ : Tuple = get_resize_output_image_size(lowerCamelCase__ , size=lowerCamelCase__ , default_to_square=lowerCamelCase__ )
else:
if "shortest_edge" in size:
a__ : int = get_resize_output_image_size(lowerCamelCase__ , size=size["shortest_edge"] , default_to_square=lowerCamelCase__ )
elif "height" in size and "width" in size:
a__ : Optional[int] = (size["height"], size["width"])
else:
raise ValueError("Invalid size for resize: {}".format(lowerCamelCase__ ) )
return resize(lowerCamelCase__ , size=lowerCamelCase__ , resample=lowerCamelCase__ , data_format=lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : np.ndarray , lowerCamelCase__ : Dict[str, int] , lowerCamelCase__ : Optional[Union[str, ChannelDimension]] = None , **lowerCamelCase__ : Any , ):
a__ : List[Any] = get_size_dict(lowerCamelCase__ )
if "height" not in size or "width" not in size:
raise ValueError(f'''size must contain \'height\' and \'width\' as keys. Got {size.keys()}''' )
return center_crop(lowerCamelCase__ , size=(size["height"], size["width"]) , data_format=lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : int , lowerCamelCase__ : np.ndarray , lowerCamelCase__ : Union[int, float] , lowerCamelCase__ : Optional[Union[str, ChannelDimension]] = None , **lowerCamelCase__ : Tuple , ):
return rescale(lowerCamelCase__ , scale=lowerCamelCase__ , data_format=lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : np.ndarray , lowerCamelCase__ : Union[float, List[float]] , lowerCamelCase__ : Union[float, List[float]] , lowerCamelCase__ : Optional[Union[str, ChannelDimension]] = None , **lowerCamelCase__ : Any , ):
return normalize(lowerCamelCase__ , mean=lowerCamelCase__ , std=lowerCamelCase__ , data_format=lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : ImageInput , lowerCamelCase__ : bool = None , lowerCamelCase__ : Dict[str, int] = None , lowerCamelCase__ : int = None , lowerCamelCase__ : PILImageResampling = None , lowerCamelCase__ : bool = None , lowerCamelCase__ : Dict[str, int] = None , lowerCamelCase__ : bool = None , lowerCamelCase__ : float = None , lowerCamelCase__ : bool = None , lowerCamelCase__ : Optional[Union[float, List[float]]] = None , lowerCamelCase__ : Optional[Union[float, List[float]]] = None , lowerCamelCase__ : Optional[Union[str, TensorType]] = None , lowerCamelCase__ : ChannelDimension = ChannelDimension.FIRST , **lowerCamelCase__ : str , ):
a__ : Union[str, Any] = do_resize if do_resize is not None else self.do_resize
a__ : Optional[Any] = crop_pct if crop_pct is not None else self.crop_pct
a__ : int = resample if resample is not None else self.resample
a__ : Any = do_center_crop if do_center_crop is not None else self.do_center_crop
a__ : int = do_rescale if do_rescale is not None else self.do_rescale
a__ : List[str] = rescale_factor if rescale_factor is not None else self.rescale_factor
a__ : int = do_normalize if do_normalize is not None else self.do_normalize
a__ : Union[str, Any] = image_mean if image_mean is not None else self.image_mean
a__ : List[Any] = image_std if image_std is not None else self.image_std
a__ : Tuple = size if size is not None else self.size
a__ : Union[str, Any] = get_size_dict(lowerCamelCase__ , default_to_square=lowerCamelCase__ )
a__ : Optional[int] = crop_size if crop_size is not None else self.crop_size
a__ : List[str] = get_size_dict(lowerCamelCase__ , param_name="crop_size" )
a__ : Union[str, Any] = make_list_of_images(lowerCamelCase__ )
if not valid_images(lowerCamelCase__ ):
raise ValueError(
"Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
"torch.Tensor, tf.Tensor or jax.ndarray." )
if do_resize and size is None or resample is None:
raise ValueError("Size and resample must be specified if do_resize is True." )
if do_center_crop and crop_pct is None:
raise ValueError("Crop_pct must be specified if do_center_crop is True." )
if do_rescale and rescale_factor is None:
raise ValueError("Rescale factor must be specified if do_rescale is True." )
if do_normalize and (image_mean is None or image_std is None):
raise ValueError("Image mean and std must be specified if do_normalize is True." )
# All transformations expect numpy arrays.
a__ : Optional[int] = [to_numpy_array(lowerCamelCase__ ) for image in images]
if do_resize:
a__ : Any = [self.resize(image=lowerCamelCase__ , size=lowerCamelCase__ , crop_pct=lowerCamelCase__ , resample=lowerCamelCase__ ) for image in images]
if do_center_crop:
a__ : Tuple = [self.center_crop(image=lowerCamelCase__ , size=lowerCamelCase__ ) for image in images]
if do_rescale:
a__ : List[Any] = [self.rescale(image=lowerCamelCase__ , scale=lowerCamelCase__ ) for image in images]
if do_normalize:
a__ : Optional[int] = [self.normalize(image=lowerCamelCase__ , mean=lowerCamelCase__ , std=lowerCamelCase__ ) for image in images]
a__ : Tuple = [to_channel_dimension_format(lowerCamelCase__ , lowerCamelCase__ ) for image in images]
a__ : Optional[Any] = {"pixel_values": images}
return BatchFeature(data=lowerCamelCase__ , tensor_type=lowerCamelCase__ )
| 37 |
import os
from typing import Optional
import fsspec
from fsspec.archive import AbstractArchiveFileSystem
from fsspec.utils import DEFAULT_BLOCK_SIZE
class A__ ( A__ ):
"""simple docstring"""
_lowercase = ''
_lowercase = (
None # protocol passed in prefix to the url. ex: "gzip", for gzip://file.txt::http://foo.bar/file.txt.gz
)
_lowercase = None # compression type in fsspec. ex: "gzip"
_lowercase = None # extension of the filename to strip. ex: "".gz" to get file.txt from file.txt.gz
def __init__( self : List[str] , lowerCamelCase__ : str = "" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , **lowerCamelCase__ : List[str] ):
super().__init__(self , **lowerCamelCase__ )
# always open as "rb" since fsspec can then use the TextIOWrapper to make it work for "r" mode
a__ : str = fsspec.open(
lowerCamelCase__ , mode="rb" , protocol=lowerCamelCase__ , compression=self.compression , client_kwargs={
"requote_redirect_url": False, # see https://github.com/huggingface/datasets/pull/5459
"trust_env": True, # Enable reading proxy env variables.
**(target_options or {}).pop("client_kwargs" , {} ), # To avoid issues if it was already passed.
} , **(target_options or {}) , )
a__ : Optional[int] = os.path.basename(self.file.path.split("::" )[0] )
a__ : int = (
self.compressed_name[: self.compressed_name.rindex("." )]
if "." in self.compressed_name
else self.compressed_name
)
a__ : List[Any] = None
@classmethod
def _UpperCamelCase( cls : int , lowerCamelCase__ : int ):
# compressed file paths are always relative to the archive root
return super()._strip_protocol(lowerCamelCase__ ).lstrip("/" )
def _UpperCamelCase( self : Dict ):
if self.dir_cache is None:
a__ : Dict = {**self.file.fs.info(self.file.path ), "name": self.uncompressed_name}
a__ : int = {f["name"]: f}
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str ):
return self.file.open().read()
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : int=None , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[str]=None , **lowerCamelCase__ : Optional[Any] , ):
a__ : Optional[int] = self._strip_protocol(lowerCamelCase__ )
if mode != "rb":
raise ValueError(f'''Tried to read with mode {mode} on file {self.file.path} opened with mode \'rb\'''' )
return self.file.open()
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'bz2'
_lowercase = 'bz2'
_lowercase = '.bz2'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gzip'
_lowercase = 'gzip'
_lowercase = '.gz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'lz4'
_lowercase = 'lz4'
_lowercase = '.lz4'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'xz'
_lowercase = 'xz'
_lowercase = '.xz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'zstd'
_lowercase = 'zstd'
_lowercase = '.zst'
def __init__( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , lowerCamelCase__ : int = DEFAULT_BLOCK_SIZE , **lowerCamelCase__ : Tuple , ):
super().__init__(
fo=lowerCamelCase__ , mode=lowerCamelCase__ , target_protocol=lowerCamelCase__ , target_options=lowerCamelCase__ , block_size=lowerCamelCase__ , **lowerCamelCase__ , )
# We need to wrap the zstd decompressor to avoid this error in fsspec==2021.7.0 and zstandard==0.15.2:
#
# File "/Users/user/.virtualenvs/hf-datasets/lib/python3.7/site-packages/fsspec/core.py", line 145, in open
# out.close = close
# AttributeError: 'zstd.ZstdDecompressionReader' object attribute 'close' is read-only
#
# see https://github.com/intake/filesystem_spec/issues/725
a__ : Any = self.file.__enter__
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : str ):
a__ : List[Any] = file_
def __enter__( self : str ):
self._file.__enter__()
return self
def __exit__( self : int , *lowerCamelCase__ : List[str] , **lowerCamelCase__ : int ):
self._file.__exit__(*lowerCamelCase__ , **lowerCamelCase__ )
def __iter__( self : List[str] ):
return iter(self._file )
def _UpperCamelCase( self : Any ):
return next(self._file )
def __getattr__( self : Optional[Any] , lowerCamelCase__ : Tuple ):
return getattr(self._file , lowerCamelCase__ )
def fixed_enter(*lowerCamelCase__ : List[str] , **lowerCamelCase__ : str ):
return WrappedFile(_enter(*lowerCamelCase__ , **lowerCamelCase__ ) )
a__ : Any = fixed_enter
| 37 | 1 |
def UpperCamelCase_ ( __a ) -> float:
return 10 - x * x
def UpperCamelCase_ ( __a , __a ) -> float:
# Bolzano theory in order to find if there is a root between a and b
if equation(__a ) * equation(__a ) >= 0:
raise ValueError("Wrong space!" )
a__ : Any = a
while (b - a) >= 0.01:
# Find middle point
a__ : str = (a + b) / 2
# Check if middle point is root
if equation(__a ) == 0.0:
break
# Decide the side to repeat the steps
if equation(__a ) * equation(__a ) < 0:
a__ : Dict = c
else:
a__ : Tuple = c
return c
if __name__ == "__main__":
import doctest
doctest.testmod()
print(bisection(-2, 5))
print(bisection(0, 6))
| 37 |
import argparse
import os
import numpy as np
import tensorflow as tf
import torch
from transformers import BertModel
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Union[str, Any] = ("dense.weight", "attention.self.query", "attention.self.key", "attention.self.value")
a__ : Union[str, Any] = (
("layer.", "layer_"),
("word_embeddings.weight", "word_embeddings"),
("position_embeddings.weight", "position_embeddings"),
("token_type_embeddings.weight", "token_type_embeddings"),
(".", "/"),
("LayerNorm/weight", "LayerNorm/gamma"),
("LayerNorm/bias", "LayerNorm/beta"),
("weight", "kernel"),
)
if not os.path.isdir(__a ):
os.makedirs(__a )
a__ : Any = model.state_dict()
def to_tf_var_name(__a ):
for patt, repl in iter(__a ):
a__ : Tuple = name.replace(__a , __a )
return f'''bert/{name}'''
def create_tf_var(__a , __a , __a ):
a__ : Tuple = tf.dtypes.as_dtype(tensor.dtype )
a__ : Dict = tf.get_variable(dtype=__a , shape=tensor.shape , name=__a , initializer=tf.zeros_initializer() )
session.run(tf.variables_initializer([tf_var] ) )
session.run(__a )
return tf_var
tf.reset_default_graph()
with tf.Session() as session:
for var_name in state_dict:
a__ : int = to_tf_var_name(__a )
a__ : Union[str, Any] = state_dict[var_name].numpy()
if any(x in var_name for x in tensors_to_transpose ):
a__ : int = torch_tensor.T
a__ : Optional[Any] = create_tf_var(tensor=__a , name=__a , session=__a )
tf.keras.backend.set_value(__a , __a )
a__ : int = session.run(__a )
print(f'''Successfully created {tf_name}: {np.allclose(__a , __a )}''' )
a__ : Any = tf.train.Saver(tf.trainable_variables() )
saver.save(__a , os.path.join(__a , model_name.replace("-" , "_" ) + ".ckpt" ) )
def UpperCamelCase_ ( __a=None ) -> int:
a__ : Dict = argparse.ArgumentParser()
parser.add_argument("--model_name" , type=__a , required=__a , help="model name e.g. bert-base-uncased" )
parser.add_argument(
"--cache_dir" , type=__a , default=__a , required=__a , help="Directory containing pytorch model" )
parser.add_argument("--pytorch_model_path" , type=__a , required=__a , help="/path/to/<pytorch-model-name>.bin" )
parser.add_argument("--tf_cache_dir" , type=__a , required=__a , help="Directory in which to save tensorflow model" )
a__ : Optional[Any] = parser.parse_args(__a )
a__ : Tuple = BertModel.from_pretrained(
pretrained_model_name_or_path=args.model_name , state_dict=torch.load(args.pytorch_model_path ) , cache_dir=args.cache_dir , )
convert_pytorch_checkpoint_to_tf(model=__a , ckpt_dir=args.tf_cache_dir , model_name=args.model_name )
if __name__ == "__main__":
main()
| 37 | 1 |
from typing import List, Optional, Tuple, Union
import torch
from ...models import UNetaDModel
from ...schedulers import KarrasVeScheduler
from ...utils import randn_tensor
from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 42
_lowercase = 42
def __init__( self : str , lowerCamelCase__ : UNetaDModel , lowerCamelCase__ : KarrasVeScheduler ):
super().__init__()
self.register_modules(unet=lowerCamelCase__ , scheduler=lowerCamelCase__ )
@torch.no_grad()
def __call__( self : Dict , lowerCamelCase__ : int = 1 , lowerCamelCase__ : int = 50 , lowerCamelCase__ : Optional[Union[torch.Generator, List[torch.Generator]]] = None , lowerCamelCase__ : Optional[str] = "pil" , lowerCamelCase__ : bool = True , **lowerCamelCase__ : Optional[int] , ):
a__ : int = self.unet.config.sample_size
a__ : Any = (batch_size, 3, img_size, img_size)
a__ : List[Any] = self.unet
# sample x_0 ~ N(0, sigma_0^2 * I)
a__ : List[Any] = randn_tensor(lowerCamelCase__ , generator=lowerCamelCase__ , device=self.device ) * self.scheduler.init_noise_sigma
self.scheduler.set_timesteps(lowerCamelCase__ )
for t in self.progress_bar(self.scheduler.timesteps ):
# here sigma_t == t_i from the paper
a__ : Dict = self.scheduler.schedule[t]
a__ : List[Any] = self.scheduler.schedule[t - 1] if t > 0 else 0
# 1. Select temporarily increased noise level sigma_hat
# 2. Add new noise to move from sample_i to sample_hat
a__, a__ : List[str] = self.scheduler.add_noise_to_input(lowerCamelCase__ , lowerCamelCase__ , generator=lowerCamelCase__ )
# 3. Predict the noise residual given the noise magnitude `sigma_hat`
# The model inputs and output are adjusted by following eq. (213) in [1].
a__ : Optional[int] = (sigma_hat / 2) * model((sample_hat + 1) / 2 , sigma_hat / 2 ).sample
# 4. Evaluate dx/dt at sigma_hat
# 5. Take Euler step from sigma to sigma_prev
a__ : Any = self.scheduler.step(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
if sigma_prev != 0:
# 6. Apply 2nd order correction
# The model inputs and output are adjusted by following eq. (213) in [1].
a__ : str = (sigma_prev / 2) * model((step_output.prev_sample + 1) / 2 , sigma_prev / 2 ).sample
a__ : Optional[int] = self.scheduler.step_correct(
lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , step_output.prev_sample , step_output["derivative"] , )
a__ : str = step_output.prev_sample
a__ : int = (sample / 2 + 0.5).clamp(0 , 1 )
a__ : Any = sample.cpu().permute(0 , 2 , 3 , 1 ).numpy()
if output_type == "pil":
a__ : str = self.numpy_to_pil(lowerCamelCase__ )
if not return_dict:
return (image,)
return ImagePipelineOutput(images=lowerCamelCase__ )
| 37 |
import inspect
import unittest
from huggingface_hub import hf_hub_download
from transformers import ASTConfig
from transformers.testing_utils import require_torch, require_torchaudio, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_torchaudio_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import ASTForAudioClassification, ASTModel
from transformers.models.audio_spectrogram_transformer.modeling_audio_spectrogram_transformer import (
AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
)
if is_torchaudio_available():
import torchaudio
from transformers import ASTFeatureExtractor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[str] , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[Any]=2 , lowerCamelCase__ : Any=24 , lowerCamelCase__ : Optional[Any]=16 , lowerCamelCase__ : int=True , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[Any]=32 , lowerCamelCase__ : List[str]=5 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Optional[Any]=37 , lowerCamelCase__ : Any="gelu" , lowerCamelCase__ : Union[str, Any]=0.1 , lowerCamelCase__ : Optional[int]=0.1 , lowerCamelCase__ : str=10 , lowerCamelCase__ : Optional[Any]=0.02 , lowerCamelCase__ : str=None , lowerCamelCase__ : List[str]=2 , lowerCamelCase__ : Optional[Any]=2 , ):
a__ : str = parent
a__ : Any = batch_size
a__ : Dict = patch_size
a__ : List[Any] = max_length
a__ : str = num_mel_bins
a__ : Optional[Any] = is_training
a__ : Optional[int] = use_labels
a__ : List[Any] = hidden_size
a__ : str = num_hidden_layers
a__ : Any = num_attention_heads
a__ : Union[str, Any] = intermediate_size
a__ : List[str] = hidden_act
a__ : str = hidden_dropout_prob
a__ : Tuple = attention_probs_dropout_prob
a__ : List[Any] = type_sequence_label_size
a__ : Any = initializer_range
a__ : str = scope
a__ : List[str] = frequency_stride
a__ : Union[str, Any] = time_stride
# in AST, the seq length equals the number of patches + 2 (we add 2 for the [CLS] and distillation tokens)
a__ : List[Any] = (self.num_mel_bins - self.patch_size) // self.frequency_stride + 1
a__ : List[str] = (self.max_length - self.patch_size) // self.time_stride + 1
a__ : Tuple = frequency_out_dimension * time_out_dimension
a__ : List[str] = num_patches + 2
def _UpperCamelCase( self : List[str] ):
a__ : Any = floats_tensor([self.batch_size, self.max_length, self.num_mel_bins] )
a__ : List[Any] = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : List[str] = self.get_config()
return config, input_values, labels
def _UpperCamelCase( self : Optional[int] ):
return ASTConfig(
patch_size=self.patch_size , max_length=self.max_length , num_mel_bins=self.num_mel_bins , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , frequency_stride=self.frequency_stride , time_stride=self.time_stride , )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : int , lowerCamelCase__ : Optional[int] ):
a__ : List[Any] = ASTModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Dict = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : str ):
a__ : Dict = self.prepare_config_and_inputs()
(
(
a__
), (
a__
), (
a__
),
) : Optional[int] = config_and_inputs
a__ : List[Any] = {"input_values": input_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(
ASTModel,
ASTForAudioClassification,
)
if is_torch_available()
else ()
)
_lowercase = (
{'audio-classification': ASTForAudioClassification, 'feature-extraction': ASTModel}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
if pipeline_test_casse_name == "AudioClassificationPipelineTests":
return True
return False
def _UpperCamelCase( self : str ):
a__ : str = ASTModelTester(self )
a__ : Any = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[str] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="AST does not use inputs_embeds" )
def _UpperCamelCase( self : List[str] ):
pass
def _UpperCamelCase( self : Optional[int] ):
a__, a__ : str = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Any = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Union[str, Any] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : Tuple ):
a__, a__ : List[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Dict = model_class(lowerCamelCase__ )
a__ : Optional[int] = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Optional[Any] = ["input_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
for model_name in AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Union[str, Any] = ASTModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : Optional[int] = hf_hub_download(
repo_id="nielsr/audio-spectogram-transformer-checkpoint" , filename="sample_audio.flac" , repo_type="dataset" )
a__, a__ : List[str] = torchaudio.load(__a )
return audio, sampling_rate
@require_torch
@require_torchaudio
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : List[str] ):
return (
ASTFeatureExtractor.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" )
if is_torchaudio_available()
else None
)
@slow
def _UpperCamelCase( self : Optional[int] ):
a__ : int = self.default_feature_extractor
a__ : Optional[Any] = ASTForAudioClassification.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" ).to(lowerCamelCase__ )
a__ : Any = self.default_feature_extractor
a__, a__ : Dict = prepare_audio()
a__ : str = audio.squeeze().numpy()
a__ : Any = feature_extractor(lowerCamelCase__ , sampling_rate=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(**lowerCamelCase__ )
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 527) )
self.assertEqual(outputs.logits.shape , lowerCamelCase__ )
a__ : List[str] = torch.tensor([-0.8760, -7.0042, -8.6602] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(outputs.logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
| 37 | 1 |
from collections import defaultdict
from graphs.minimum_spanning_tree_prims import prisms_algorithm as mst
def UpperCamelCase_ ( ) -> Optional[Any]:
a__, a__ : Tuple = 9, 14 # noqa: F841
a__ : int = [
[0, 1, 4],
[0, 7, 8],
[1, 2, 8],
[7, 8, 7],
[7, 6, 1],
[2, 8, 2],
[8, 6, 6],
[2, 3, 7],
[2, 5, 4],
[6, 5, 2],
[3, 5, 14],
[3, 4, 9],
[5, 4, 10],
[1, 7, 11],
]
a__ : str = defaultdict(__a )
for nodea, nodea, cost in edges:
adjancency[nodea].append([nodea, cost] )
adjancency[nodea].append([nodea, cost] )
a__ : Union[str, Any] = mst(__a )
a__ : str = [
[7, 6, 1],
[2, 8, 2],
[6, 5, 2],
[0, 1, 4],
[2, 5, 4],
[2, 3, 7],
[0, 7, 8],
[3, 4, 9],
]
for answer in expected:
a__ : List[str] = tuple(answer[:2] )
a__ : Any = tuple(edge[::-1] )
assert edge in result or reverse in result
| 37 |
import pickle
import shutil
import tempfile
import unittest
from transformers import SPIECE_UNDERLINE, XGLMTokenizer, XGLMTokenizerFast
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers, slow
from transformers.utils import cached_property
from ...test_tokenization_common import TokenizerTesterMixin
UpperCamelCase : Optional[Any] = get_tests_dir("""fixtures/test_sentencepiece.model""")
@require_sentencepiece
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = XGLMTokenizer
_lowercase = XGLMTokenizerFast
_lowercase = True
_lowercase = True
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# We have a SentencePiece fixture for testing
a__ : str = XGLMTokenizer(lowerCamelCase__ , keep_accents=lowerCamelCase__ )
tokenizer.save_pretrained(self.tmpdirname )
def _UpperCamelCase( self : List[Any] ):
a__ : int = "<pad>"
a__ : Union[str, Any] = 1
self.assertEqual(self.get_tokenizer()._convert_token_to_id(lowerCamelCase__ ) , lowerCamelCase__ )
self.assertEqual(self.get_tokenizer()._convert_id_to_token(lowerCamelCase__ ) , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : List[str] = list(self.get_tokenizer().get_vocab().keys() )
self.assertEqual(vocab_keys[0] , "<s>" )
self.assertEqual(vocab_keys[1] , "<pad>" )
self.assertEqual(len(lowerCamelCase__ ) , 1_008 )
def _UpperCamelCase( self : Dict ):
self.assertEqual(self.get_tokenizer().vocab_size , 1_008 )
def _UpperCamelCase( self : Optional[int] ):
a__ : str = XGLMTokenizer(lowerCamelCase__ , keep_accents=lowerCamelCase__ )
a__ : List[str] = tokenizer.tokenize("This is a test" )
self.assertListEqual(lowerCamelCase__ , ["▁This", "▁is", "▁a", "▁t", "est"] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
a__ : Any = tokenizer.tokenize("I was born in 92000, and this is falsé." )
self.assertListEqual(
lowerCamelCase__ , [
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"9",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"é",
".",
] , )
a__ : List[str] = tokenizer.convert_tokens_to_ids(lowerCamelCase__ )
self.assertListEqual(
lowerCamelCase__ , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
a__ : Dict = tokenizer.convert_ids_to_tokens(lowerCamelCase__ )
self.assertListEqual(
lowerCamelCase__ , [
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"<unk>",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"<unk>",
".",
] , )
@cached_property
def _UpperCamelCase( self : Dict ):
return XGLMTokenizer.from_pretrained("facebook/xglm-564M" )
def _UpperCamelCase( self : Union[str, Any] ):
with tempfile.NamedTemporaryFile() as f:
shutil.copyfile(lowerCamelCase__ , f.name )
a__ : Any = XGLMTokenizer(f.name , keep_accents=lowerCamelCase__ )
a__ : List[str] = pickle.dumps(lowerCamelCase__ )
pickle.loads(lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
if not self.test_rust_tokenizer:
return
a__ : Any = self.get_tokenizer()
a__ : Optional[Any] = self.get_rust_tokenizer()
a__ : Tuple = "I was born in 92000, and this is falsé."
a__ : List[str] = tokenizer.tokenize(lowerCamelCase__ )
a__ : Union[str, Any] = rust_tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : Optional[int] = tokenizer.encode(lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
a__ : Union[str, Any] = rust_tokenizer.encode(lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : List[str] = self.get_rust_tokenizer()
a__ : Tuple = tokenizer.encode(lowerCamelCase__ )
a__ : Optional[Any] = rust_tokenizer.encode(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = "Hello World!"
a__ : List[str] = [2, 31_227, 4_447, 35]
self.assertListEqual(lowerCamelCase__ , self.big_tokenizer.encode(lowerCamelCase__ ) )
@slow
def _UpperCamelCase( self : Union[str, Any] ):
a__ : Optional[int] = (
"This is a very long text with a lot of weird characters, such as: . , ~ ? ( ) \" [ ] ! : - . Also we will"
" add words that should not exsist and be tokenized to unk, such as saoneuhaoesuth"
)
# fmt: off
a__ : Union[str, Any] = [2, 1_018, 67, 11, 1_988, 2_617, 5_631, 278, 11, 3_407, 48, 71_630, 28_085, 4, 3_234, 157, 13, 6, 5, 6, 4, 3_526, 768, 15, 659, 57, 298, 3_983, 864, 129, 21, 6, 5, 13_675, 377, 652, 7_580, 10_341, 155, 2_817, 422, 1_666, 7, 1_674, 53, 113, 202_277, 17_892, 33, 60, 87, 4, 3_234, 157, 61, 2_667, 52_376, 19, 88, 23, 735]
# fmt: on
self.assertListEqual(lowerCamelCase__ , self.big_tokenizer.encode(lowerCamelCase__ ) )
@slow
def _UpperCamelCase( self : List[Any] ):
# fmt: off
a__ : Optional[int] = {
"input_ids": [[2, 108_825, 1_163, 15, 88_010, 473, 15_898, 157, 13_672, 1_857, 312, 8, 238_021, 1_163, 53, 13_672, 1_857, 312, 8, 53_283, 182_396, 8, 18_566, 16, 36_733, 4_101, 8, 230, 244_017, 122_553, 7, 15, 132_597, 4, 293, 12_511, 7_610, 4, 3_414, 132_597, 9, 4, 32_361, 362, 4, 734, 28_512, 32_569, 18, 4, 32_361, 26_096, 14_982, 73, 18_715, 21_433, 235_261, 15, 492, 12_427, 16, 53, 18_715, 21_433, 65_454, 15, 23_659, 563, 16, 278, 597, 2_843, 595, 7_931, 182_396, 64_186, 22, 886, 595, 132_981, 53, 25_540, 3_449, 43_982, 39_901, 5_951, 878, 330, 4, 27_694, 80_269, 312, 53, 6_517, 11_780, 611, 20_408, 5], [2, 6, 132_597, 67, 42_897, 33, 592, 8, 163_729, 25_540, 361, 136_997, 109_514, 173_230, 7, 501, 60, 102_913, 196, 5_631, 235, 63_243, 473, 6, 231_757, 74, 5_277, 7_905, 53, 3_095, 37_317, 22, 454, 183_874, 5], [2, 268, 31_298, 46_530, 6, 132_935, 43_831, 7, 597, 32, 24, 3_688, 9_865, 5]],
"attention_mask": [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]
} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=lowerCamelCase__ , model_name="facebook/xglm-564M" , padding=lowerCamelCase__ , )
| 37 | 1 |
from __future__ import annotations
def UpperCamelCase_ ( __a , __a ) -> list[int]:
a__ : Dict = 0
a__ : Any = len(__a ) - 1
while i < j:
if nums[i] + nums[j] == target:
return [i, j]
elif nums[i] + nums[j] < target:
a__ : Optional[int] = i + 1
else:
a__ : Optional[Any] = j - 1
return []
if __name__ == "__main__":
import doctest
doctest.testmod()
print(f"""{two_pointer([2, 7, 11, 15], 9) = }""")
| 37 |
import argparse
import requests
import torch
# pip3 install salesforce-lavis
# I'm actually installing a slightly modified version: pip3 install git+https://github.com/nielsrogge/LAVIS.git@fix_lavis_float32 (there's also the fix_lavis branch)
# also note: to convert Vicuna checkpoints, we had to include /home/niels/python_projects/checkpoints/FastChat/vicuna-7b in lavis/configs/models/blip2/blip2_instruct_vicuna7b.yaml
# same for Vicuna-13b
from lavis.models import load_model_and_preprocess
from PIL import Image
from transformers import (
AutoTokenizer,
BlipImageProcessor,
InstructBlipConfig,
InstructBlipForConditionalGeneration,
InstructBlipProcessor,
InstructBlipQFormerConfig,
InstructBlipVisionConfig,
LlamaConfig,
LlamaTokenizerFast,
TaConfig,
TaTokenizerFast,
)
from transformers.utils.constants import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD
def UpperCamelCase_ ( ) -> int:
a__ : int = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
a__ : Optional[Any] = Image.open(requests.get(__a , stream=__a ).raw ).convert("RGB" )
return image
def UpperCamelCase_ ( __a ) -> Optional[Any]:
a__ : Any = []
# fmt: off
# vision encoder
rename_keys.append(("visual_encoder.cls_token", "vision_model.embeddings.class_embedding") )
rename_keys.append(("visual_encoder.pos_embed", "vision_model.embeddings.position_embedding") )
rename_keys.append(("visual_encoder.patch_embed.proj.weight", "vision_model.embeddings.patch_embedding.weight") )
rename_keys.append(("visual_encoder.patch_embed.proj.bias", "vision_model.embeddings.patch_embedding.bias") )
rename_keys.append(("ln_vision.weight", "vision_model.post_layernorm.weight") )
rename_keys.append(("ln_vision.bias", "vision_model.post_layernorm.bias") )
for i in range(config.vision_config.num_hidden_layers ):
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.weight''', f'''vision_model.encoder.layers.{i}.layer_norm1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.bias''', f'''vision_model.encoder.layers.{i}.layer_norm1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.weight''', f'''vision_model.encoder.layers.{i}.layer_norm2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.bias''', f'''vision_model.encoder.layers.{i}.layer_norm2.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.qkv.weight''', f'''vision_model.encoder.layers.{i}.self_attn.qkv.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.weight''', f'''vision_model.encoder.layers.{i}.self_attn.projection.weight''',) )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.bias''', f'''vision_model.encoder.layers.{i}.self_attn.projection.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc2.bias''') )
# QFormer
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.weight", "qformer.embeddings.layernorm.weight") )
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.bias", "qformer.embeddings.layernorm.bias") )
# fmt: on
return rename_keys
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Union[str, Any] = dct.pop(__a )
a__ : List[str] = val
def UpperCamelCase_ ( __a , __a ) -> Optional[Any]:
for i in range(config.vision_config.num_hidden_layers ):
# read in original q and v biases
a__ : Any = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.q_bias''' )
a__ : Tuple = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.v_bias''' )
# next, set bias in the state dict
a__ : str = torch.cat((q_bias, torch.zeros_like(__a , requires_grad=__a ), v_bias) )
a__ : int = qkv_bias
def UpperCamelCase_ ( __a ) -> Dict:
a__ : Tuple = 364 if "coco" in model_name else 224
a__ : int = InstructBlipVisionConfig(image_size=__a ).to_dict()
# make sure the models have proper bos_token_id and eos_token_id set (important for generation)
# seems like flan-T5 models don't have bos_token_id properly set?
if "t5-xl" in model_name:
a__ : Tuple = TaConfig.from_pretrained("google/flan-t5-xl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "t5-xxl" in model_name:
a__ : Dict = TaConfig.from_pretrained("google/flan-t5-xxl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "vicuna-7b" in model_name:
a__ : List[Any] = LlamaConfig.from_pretrained("decapoda-research/llama-7b-hf" , vocab_size=32_001 ).to_dict()
elif "vicuna-13b" in model_name:
a__ : Optional[int] = LlamaConfig.from_pretrained("decapoda-research/llama-13b-hf" , vocab_size=32_001 ).to_dict()
else:
raise ValueError("Model name not supported" )
# the authors add one special "[DEC]" token to the vocab of Q-Former, hence vocab size = 30522 + 1
a__ : Optional[Any] = InstructBlipQFormerConfig(vocab_size=30_523 ).to_dict()
a__ : Any = InstructBlipConfig(vision_config=__a , text_config=__a , qformer_config=__a )
return config, image_size
@torch.no_grad()
def UpperCamelCase_ ( __a , __a=None , __a=False ) -> int:
a__ : Tuple = AutoTokenizer.from_pretrained("bert-base-uncased" , truncation_side="left" )
qformer_tokenizer.add_special_tokens({"bos_token": "[DEC]"} )
if "t5" in model_name:
a__ : List[Any] = TaTokenizerFast.from_pretrained("google/flan-t5-xl" , truncation_side="left" )
elif "vicuna" in model_name:
# the following was used in the original implementation:
# tokenizer = LlamaTokenizer.from_pretrained("huggyllama/llama-7b", use_fast=False, truncation_side="left")
# tokenizer.add_special_tokens({"pad_token": "[PAD]"})
# tokenizer.add_special_tokens({"bos_token": "</s>"})
# tokenizer.add_special_tokens({"eos_token": "</s>"})
# tokenizer.add_special_tokens({"unk_token": "</s>"})
a__ : Union[str, Any] = LlamaTokenizerFast.from_pretrained(
"huggyllama/llama-7b" , truncation_side="left" , bos_token="</s>" , unk_token="</s>" )
tokenizer.add_special_tokens({"pad_token": "[PAD]"} )
a__, a__ : List[str] = get_blipa_config(__a )
a__ : Any = InstructBlipForConditionalGeneration(__a ).eval()
a__ : Dict = {
"instructblip-vicuna-7b": ("blip2_vicuna_instruct", "vicuna7b"),
"instructblip-vicuna-13b": ("blip2_vicuna_instruct", "vicuna13b"),
"instructblip-flan-t5-xl": ("blip2_t5_instruct", "flant5xl"),
"instructblip-flan-t5-xxl": ("blip2_t5_instruct", "flant5xxl"),
}
a__, a__ : Dict = model_name_to_original[model_name]
# load original model
print("Loading original model..." )
a__ : Optional[Any] = "cuda:1" if torch.cuda.is_available() else "cpu"
a__ : List[Any] = "cuda:2" if torch.cuda.is_available() else "cpu"
a__, a__, a__ : Tuple = load_model_and_preprocess(
name=__a , model_type=__a , is_eval=__a , device=__a )
original_model.eval()
print("Done!" )
# update state dict keys
a__ : Dict = original_model.state_dict()
a__ : Optional[int] = create_rename_keys(__a )
for src, dest in rename_keys:
rename_key(__a , __a , __a )
# some keys can be renamed efficiently
for key, val in state_dict.copy().items():
a__ : Optional[int] = state_dict.pop(__a )
if key.startswith("Qformer.bert" ):
a__ : List[Any] = key.replace("Qformer.bert" , "qformer" )
if "attention.self" in key:
a__ : Any = key.replace("self" , "attention" )
if "llm_proj" in key:
a__ : Dict = key.replace("llm_proj" , "language_projection" )
if "t5_proj" in key:
a__ : int = key.replace("t5_proj" , "language_projection" )
if key.startswith("llm_model" ):
a__ : List[str] = key.replace("llm_model" , "language_model" )
if key.startswith("t5" ):
a__ : str = key.replace("t5" , "language" )
a__ : Dict = val
# read in qv biases
read_in_q_v_bias(__a , __a )
# note: weights get loaded in torch.float32 by default
hf_model.load_state_dict(__a , strict=__a )
a__ : Union[str, Any] = load_demo_image()
a__ : int = "What is unusual about this image?"
# create processor
a__ : Any = BlipImageProcessor(
size={"height": image_size, "width": image_size} , image_mean=__a , image_std=__a )
a__ : Tuple = InstructBlipProcessor(
image_processor=__a , tokenizer=__a , qformer_tokenizer=__a , )
a__ : Tuple = processor(images=__a , text=__a , return_tensors="pt" ).to(__a )
# make sure processor creates exact same pixel values
a__ : Optional[int] = vis_processors["eval"](__a ).unsqueeze(0 ).to(__a )
a__ : Optional[Any] = inputs.pixel_values
assert torch.allclose(original_pixel_values.to(pixel_values.device ) , __a )
original_model.to(__a )
hf_model.to(__a )
with torch.no_grad():
if "vicuna" in model_name:
a__ : str = original_model({"image": original_pixel_values, "text_input": [prompt]} ).logits
a__ : List[str] = hf_model(**__a ).logits
else:
a__ : List[Any] = original_model(
{"image": original_pixel_values, "text_input": [prompt], "text_output": ["\n"]} ).logits
a__ : str = tokenizer("\n" , return_tensors="pt" ).input_ids.to(__a )
a__ : Dict = label_input_ids.masked_fill(label_input_ids == tokenizer.pad_token_id , -100 )
a__ : Any = hf_model(**__a , labels=__a ).logits
print("First values of original logits:" , original_logits[0, :3, :3] )
print("First values of HF logits:" , logits[0, :3, :3] )
# assert values
assert original_logits.shape == logits.shape
a__ : Tuple = 1e-4 if "vicuna" in model_name else 1e-5
assert torch.allclose(original_logits.to(logits.device ) , __a , atol=__a )
print("Looks ok!" )
print("Generating with original model..." )
a__ : Tuple = original_model.generate({"image": original_pixel_values, "prompt": prompt} , num_beams=5 )
# important: we need to cast the weights of the HF model to the appropriate type
print("Generating with HF model..." )
a__ : int = hf_model.generate(
**__a , do_sample=__a , num_beams=5 , max_length=256 , min_length=1 , top_p=0.9 , repetition_penalty=1.5 , length_penalty=1.0 , temperature=1 , )
if "vicuna" in model_name:
# convert output id 0 to 2 (eos_token_id)
# TODO add this in the generate method?
a__ : int = 2
print("Original generation:" , __a )
a__ : str = processor.batch_decode(__a , skip_special_tokens=__a )
a__ : str = [text.strip() for text in output_text]
print("HF generation:" , __a )
if pytorch_dump_folder_path is not None:
processor.save_pretrained(__a )
hf_model.save_pretrained(__a )
if push_to_hub:
processor.push_to_hub(f'''Salesforce/{model_name}''' )
hf_model.push_to_hub(f'''Salesforce/{model_name}''' )
if __name__ == "__main__":
UpperCamelCase : Any = argparse.ArgumentParser()
UpperCamelCase : Optional[int] = [
"""instructblip-vicuna-7b""",
"""instructblip-vicuna-13b""",
"""instructblip-flan-t5-xl""",
"""instructblip-flan-t5-xxl""",
]
parser.add_argument(
"""--model_name""",
default="""instructblip-flan-t5-xl""",
choices=choices,
type=str,
help="""Path to hf config.json of model to convert""",
)
parser.add_argument("""--pytorch_dump_folder_path""", default=None, type=str, help="""Path to the output PyTorch model.""")
parser.add_argument(
"""--push_to_hub""",
action="""store_true""",
help="""Whether to push the model and processor to the hub after converting""",
)
UpperCamelCase : Dict = parser.parse_args()
convert_blipa_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 37 | 1 |
import unittest
import numpy as np
from transformers.testing_utils import require_pytesseract, require_torch
from transformers.utils import is_pytesseract_available, is_torch_available
from ...test_image_processing_common import ImageProcessingSavingTestMixin, prepare_image_inputs
if is_torch_available():
import torch
if is_pytesseract_available():
from PIL import Image
from transformers import LayoutLMvaImageProcessor
class A__ ( unittest.TestCase ):
"""simple docstring"""
def __init__( self : Tuple , lowerCamelCase__ : Any , lowerCamelCase__ : Dict=7 , lowerCamelCase__ : int=3 , lowerCamelCase__ : Dict=18 , lowerCamelCase__ : List[str]=30 , lowerCamelCase__ : int=400 , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : Tuple=None , lowerCamelCase__ : List[Any]=True , ):
a__ : Dict = size if size is not None else {"height": 18, "width": 18}
a__ : Union[str, Any] = parent
a__ : Dict = batch_size
a__ : Union[str, Any] = num_channels
a__ : Optional[int] = image_size
a__ : Tuple = min_resolution
a__ : Dict = max_resolution
a__ : Tuple = do_resize
a__ : int = size
a__ : List[str] = apply_ocr
def _UpperCamelCase( self : List[Any] ):
return {"do_resize": self.do_resize, "size": self.size, "apply_ocr": self.apply_ocr}
@require_torch
@require_pytesseract
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = LayoutLMvaImageProcessor if is_pytesseract_available() else None
def _UpperCamelCase( self : Dict ):
a__ : Any = LayoutLMvaImageProcessingTester(self )
@property
def _UpperCamelCase( self : Optional[Any] ):
return self.image_processor_tester.prepare_image_processor_dict()
def _UpperCamelCase( self : Optional[int] ):
a__ : Tuple = self.image_processing_class(**self.image_processor_dict )
self.assertTrue(hasattr(lowerCamelCase__ , "do_resize" ) )
self.assertTrue(hasattr(lowerCamelCase__ , "size" ) )
self.assertTrue(hasattr(lowerCamelCase__ , "apply_ocr" ) )
def _UpperCamelCase( self : Any ):
a__ : int = self.image_processing_class.from_dict(self.image_processor_dict )
self.assertEqual(image_processor.size , {"height": 18, "width": 18} )
a__ : Optional[Any] = self.image_processing_class.from_dict(self.image_processor_dict , size=42 )
self.assertEqual(image_processor.size , {"height": 42, "width": 42} )
def _UpperCamelCase( self : Dict ):
pass
def _UpperCamelCase( self : Optional[Any] ):
# Initialize image_processing
a__ : Dict = self.image_processing_class(**self.image_processor_dict )
# create random PIL images
a__ : Tuple = prepare_image_inputs(self.image_processor_tester , equal_resolution=lowerCamelCase__ )
for image in image_inputs:
self.assertIsInstance(lowerCamelCase__ , Image.Image )
# Test not batched input
a__ : List[str] = image_processing(image_inputs[0] , return_tensors="pt" )
self.assertEqual(
encoding.pixel_values.shape , (
1,
self.image_processor_tester.num_channels,
self.image_processor_tester.size["height"],
self.image_processor_tester.size["width"],
) , )
self.assertIsInstance(encoding.words , lowerCamelCase__ )
self.assertIsInstance(encoding.boxes , lowerCamelCase__ )
# Test batched
a__ : Union[str, Any] = image_processing(lowerCamelCase__ , return_tensors="pt" ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_processor_tester.batch_size,
self.image_processor_tester.num_channels,
self.image_processor_tester.size["height"],
self.image_processor_tester.size["width"],
) , )
def _UpperCamelCase( self : int ):
# Initialize image_processing
a__ : Tuple = self.image_processing_class(**self.image_processor_dict )
# create random numpy tensors
a__ : Dict = prepare_image_inputs(self.image_processor_tester , equal_resolution=lowerCamelCase__ , numpify=lowerCamelCase__ )
for image in image_inputs:
self.assertIsInstance(lowerCamelCase__ , np.ndarray )
# Test not batched input
a__ : Any = image_processing(image_inputs[0] , return_tensors="pt" ).pixel_values
self.assertEqual(
encoded_images.shape , (
1,
self.image_processor_tester.num_channels,
self.image_processor_tester.size["height"],
self.image_processor_tester.size["width"],
) , )
# Test batched
a__ : Union[str, Any] = image_processing(lowerCamelCase__ , return_tensors="pt" ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_processor_tester.batch_size,
self.image_processor_tester.num_channels,
self.image_processor_tester.size["height"],
self.image_processor_tester.size["width"],
) , )
def _UpperCamelCase( self : Optional[Any] ):
# Initialize image_processing
a__ : List[str] = self.image_processing_class(**self.image_processor_dict )
# create random PyTorch tensors
a__ : Dict = prepare_image_inputs(self.image_processor_tester , equal_resolution=lowerCamelCase__ , torchify=lowerCamelCase__ )
for image in image_inputs:
self.assertIsInstance(lowerCamelCase__ , torch.Tensor )
# Test not batched input
a__ : List[Any] = image_processing(image_inputs[0] , return_tensors="pt" ).pixel_values
self.assertEqual(
encoded_images.shape , (
1,
self.image_processor_tester.num_channels,
self.image_processor_tester.size["height"],
self.image_processor_tester.size["width"],
) , )
# Test batched
a__ : int = image_processing(lowerCamelCase__ , return_tensors="pt" ).pixel_values
self.assertEqual(
encoded_images.shape , (
self.image_processor_tester.batch_size,
self.image_processor_tester.num_channels,
self.image_processor_tester.size["height"],
self.image_processor_tester.size["width"],
) , )
def _UpperCamelCase( self : Union[str, Any] ):
# with apply_OCR = True
a__ : Any = LayoutLMvaImageProcessor()
from datasets import load_dataset
a__ : int = load_dataset("hf-internal-testing/fixtures_docvqa" , split="test" )
a__ : List[str] = Image.open(ds[0]["file"] ).convert("RGB" )
a__ : Any = image_processing(lowerCamelCase__ , return_tensors="pt" )
self.assertEqual(encoding.pixel_values.shape , (1, 3, 224, 224) )
self.assertEqual(len(encoding.words ) , len(encoding.boxes ) )
# fmt: off
# the words and boxes were obtained with Tesseract 4.1.1
a__ : Tuple = [["11:14", "to", "11:39", "a.m", "11:39", "to", "11:44", "a.m.", "11:44", "a.m.", "to", "12:25", "p.m.", "12:25", "to", "12:58", "p.m.", "12:58", "to", "4:00", "p.m.", "2:00", "to", "5:00", "p.m.", "Coffee", "Break", "Coffee", "will", "be", "served", "for", "men", "and", "women", "in", "the", "lobby", "adjacent", "to", "exhibit", "area.", "Please", "move", "into", "exhibit", "area.", "(Exhibits", "Open)", "TRRF", "GENERAL", "SESSION", "(PART", "|)", "Presiding:", "Lee", "A.", "Waller", "TRRF", "Vice", "President", "“Introductory", "Remarks”", "Lee", "A.", "Waller,", "TRRF", "Vice", "Presi-", "dent", "Individual", "Interviews", "with", "TRRF", "Public", "Board", "Members", "and", "Sci-", "entific", "Advisory", "Council", "Mem-", "bers", "Conducted", "by", "TRRF", "Treasurer", "Philip", "G.", "Kuehn", "to", "get", "answers", "which", "the", "public", "refrigerated", "warehousing", "industry", "is", "looking", "for.", "Plus", "questions", "from", "the", "floor.", "Dr.", "Emil", "M.", "Mrak,", "University", "of", "Cal-", "ifornia,", "Chairman,", "TRRF", "Board;", "Sam", "R.", "Cecil,", "University", "of", "Georgia", "College", "of", "Agriculture;", "Dr.", "Stanley", "Charm,", "Tufts", "University", "School", "of", "Medicine;", "Dr.", "Robert", "H.", "Cotton,", "ITT", "Continental", "Baking", "Company;", "Dr.", "Owen", "Fennema,", "University", "of", "Wis-", "consin;", "Dr.", "Robert", "E.", "Hardenburg,", "USDA.", "Questions", "and", "Answers", "Exhibits", "Open", "Capt.", "Jack", "Stoney", "Room", "TRRF", "Scientific", "Advisory", "Council", "Meeting", "Ballroom", "Foyer"]] # noqa: E231
a__ : Union[str, Any] = [[[141, 57, 214, 69], [228, 58, 252, 69], [141, 75, 216, 88], [230, 79, 280, 88], [142, 260, 218, 273], [230, 261, 255, 273], [143, 279, 218, 290], [231, 282, 290, 291], [143, 342, 218, 354], [231, 345, 289, 355], [202, 362, 227, 373], [143, 379, 220, 392], [231, 382, 291, 394], [144, 714, 220, 726], [231, 715, 256, 726], [144, 732, 220, 745], [232, 736, 291, 747], [144, 769, 218, 782], [231, 770, 256, 782], [141, 788, 202, 801], [215, 791, 274, 804], [143, 826, 204, 838], [215, 826, 240, 838], [142, 844, 202, 857], [215, 847, 274, 859], [334, 57, 427, 69], [440, 57, 522, 69], [369, 75, 461, 88], [469, 75, 516, 88], [528, 76, 562, 88], [570, 76, 667, 88], [675, 75, 711, 87], [721, 79, 778, 88], [789, 75, 840, 88], [369, 97, 470, 107], [484, 94, 507, 106], [518, 94, 562, 107], [576, 94, 655, 110], [668, 94, 792, 109], [804, 95, 829, 107], [369, 113, 465, 125], [477, 116, 547, 125], [562, 113, 658, 125], [671, 116, 748, 125], [761, 113, 811, 125], [369, 131, 465, 143], [477, 133, 548, 143], [563, 130, 698, 145], [710, 130, 802, 146], [336, 171, 412, 183], [423, 171, 572, 183], [582, 170, 716, 184], [728, 171, 817, 187], [829, 171, 844, 186], [338, 197, 482, 212], [507, 196, 557, 209], [569, 196, 595, 208], [610, 196, 702, 209], [505, 214, 583, 226], [595, 214, 656, 227], [670, 215, 807, 227], [335, 259, 543, 274], [556, 259, 708, 272], [372, 279, 422, 291], [435, 279, 460, 291], [474, 279, 574, 292], [587, 278, 664, 291], [676, 278, 738, 291], [751, 279, 834, 291], [372, 298, 434, 310], [335, 341, 483, 354], [497, 341, 655, 354], [667, 341, 728, 354], [740, 341, 825, 354], [335, 360, 430, 372], [442, 360, 534, 372], [545, 359, 687, 372], [697, 360, 754, 372], [765, 360, 823, 373], [334, 378, 428, 391], [440, 378, 577, 394], [590, 378, 705, 391], [720, 378, 801, 391], [334, 397, 400, 409], [370, 416, 529, 429], [544, 416, 576, 432], [587, 416, 665, 428], [677, 416, 814, 429], [372, 435, 452, 450], [465, 434, 495, 447], [511, 434, 600, 447], [611, 436, 637, 447], [649, 436, 694, 451], [705, 438, 824, 447], [369, 453, 452, 466], [464, 454, 509, 466], [522, 453, 611, 469], [625, 453, 792, 469], [370, 472, 556, 488], [570, 472, 684, 487], [697, 472, 718, 485], [732, 472, 835, 488], [369, 490, 411, 503], [425, 490, 484, 503], [496, 490, 635, 506], [645, 490, 707, 503], [718, 491, 761, 503], [771, 490, 840, 503], [336, 510, 374, 521], [388, 510, 447, 522], [460, 510, 489, 521], [503, 510, 580, 522], [592, 509, 736, 525], [745, 509, 770, 522], [781, 509, 840, 522], [338, 528, 434, 541], [448, 528, 596, 541], [609, 527, 687, 540], [700, 528, 792, 541], [336, 546, 397, 559], [407, 546, 431, 559], [443, 546, 525, 560], [537, 546, 680, 562], [688, 546, 714, 559], [722, 546, 837, 562], [336, 565, 449, 581], [461, 565, 485, 577], [497, 565, 665, 581], [681, 565, 718, 577], [732, 565, 837, 580], [337, 584, 438, 597], [452, 583, 521, 596], [535, 584, 677, 599], [690, 583, 787, 596], [801, 583, 825, 596], [338, 602, 478, 615], [492, 602, 530, 614], [543, 602, 638, 615], [650, 602, 676, 614], [688, 602, 788, 615], [802, 602, 843, 614], [337, 621, 502, 633], [516, 621, 615, 637], [629, 621, 774, 636], [789, 621, 827, 633], [337, 639, 418, 652], [432, 640, 571, 653], [587, 639, 731, 655], [743, 639, 769, 652], [780, 639, 841, 652], [338, 658, 440, 673], [455, 658, 491, 670], [508, 658, 602, 671], [616, 658, 638, 670], [654, 658, 835, 674], [337, 677, 429, 689], [337, 714, 482, 726], [495, 714, 548, 726], [561, 714, 683, 726], [338, 770, 461, 782], [474, 769, 554, 785], [489, 788, 562, 803], [576, 788, 643, 801], [656, 787, 751, 804], [764, 788, 844, 801], [334, 825, 421, 838], [430, 824, 574, 838], [584, 824, 723, 841], [335, 844, 450, 857], [464, 843, 583, 860], [628, 862, 755, 875], [769, 861, 848, 878]]] # noqa: E231
# fmt: on
self.assertListEqual(encoding.words , lowerCamelCase__ )
self.assertListEqual(encoding.boxes , lowerCamelCase__ )
# with apply_OCR = False
a__ : List[str] = LayoutLMvaImageProcessor(apply_ocr=lowerCamelCase__ )
a__ : List[Any] = image_processing(lowerCamelCase__ , return_tensors="pt" )
self.assertEqual(encoding.pixel_values.shape , (1, 3, 224, 224) )
| 37 |
def UpperCamelCase_ ( __a , __a ) -> Tuple:
a__ : Optional[int] = [0 for i in range(r + 1 )]
# nc0 = 1
a__ : Union[str, Any] = 1
for i in range(1 , n + 1 ):
# to compute current row from previous row.
a__ : Any = min(__a , __a )
while j > 0:
c[j] += c[j - 1]
j -= 1
return c[r]
print(binomial_coefficient(n=10, r=5))
| 37 | 1 |
from collections import OrderedDict
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxConfig
from ...utils import logging
UpperCamelCase : Tuple = logging.get_logger(__name__)
UpperCamelCase : str = {
"""xlm-mlm-en-2048""": """https://huggingface.co/xlm-mlm-en-2048/resolve/main/config.json""",
"""xlm-mlm-ende-1024""": """https://huggingface.co/xlm-mlm-ende-1024/resolve/main/config.json""",
"""xlm-mlm-enfr-1024""": """https://huggingface.co/xlm-mlm-enfr-1024/resolve/main/config.json""",
"""xlm-mlm-enro-1024""": """https://huggingface.co/xlm-mlm-enro-1024/resolve/main/config.json""",
"""xlm-mlm-tlm-xnli15-1024""": """https://huggingface.co/xlm-mlm-tlm-xnli15-1024/resolve/main/config.json""",
"""xlm-mlm-xnli15-1024""": """https://huggingface.co/xlm-mlm-xnli15-1024/resolve/main/config.json""",
"""xlm-clm-enfr-1024""": """https://huggingface.co/xlm-clm-enfr-1024/resolve/main/config.json""",
"""xlm-clm-ende-1024""": """https://huggingface.co/xlm-clm-ende-1024/resolve/main/config.json""",
"""xlm-mlm-17-1280""": """https://huggingface.co/xlm-mlm-17-1280/resolve/main/config.json""",
"""xlm-mlm-100-1280""": """https://huggingface.co/xlm-mlm-100-1280/resolve/main/config.json""",
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'xlm'
_lowercase = {
'hidden_size': 'emb_dim',
'num_attention_heads': 'n_heads',
'num_hidden_layers': 'n_layers',
'n_words': 'vocab_size', # For backward compatibility
}
def __init__( self : Tuple , lowerCamelCase__ : List[Any]=30_145 , lowerCamelCase__ : Any=2_048 , lowerCamelCase__ : int=12 , lowerCamelCase__ : Dict=16 , lowerCamelCase__ : Optional[int]=0.1 , lowerCamelCase__ : Any=0.1 , lowerCamelCase__ : Union[str, Any]=True , lowerCamelCase__ : Dict=False , lowerCamelCase__ : Optional[int]=False , lowerCamelCase__ : List[Any]=False , lowerCamelCase__ : List[str]=1 , lowerCamelCase__ : str=True , lowerCamelCase__ : str=512 , lowerCamelCase__ : List[str]=2_048**-0.5 , lowerCamelCase__ : List[str]=1E-12 , lowerCamelCase__ : Any=0.02 , lowerCamelCase__ : Tuple=0 , lowerCamelCase__ : Any=1 , lowerCamelCase__ : Optional[Any]=2 , lowerCamelCase__ : Optional[Any]=3 , lowerCamelCase__ : Optional[Any]=5 , lowerCamelCase__ : Tuple=True , lowerCamelCase__ : Optional[Any]="first" , lowerCamelCase__ : Optional[int]=True , lowerCamelCase__ : Tuple=None , lowerCamelCase__ : int=True , lowerCamelCase__ : Dict=0.1 , lowerCamelCase__ : Tuple=5 , lowerCamelCase__ : Any=5 , lowerCamelCase__ : List[Any]=0 , lowerCamelCase__ : Optional[int]=0 , lowerCamelCase__ : Optional[Any]=2 , lowerCamelCase__ : Tuple=0 , **lowerCamelCase__ : List[str] , ):
a__ : List[str] = vocab_size
a__ : str = emb_dim
a__ : List[str] = n_layers
a__ : Union[str, Any] = n_heads
a__ : Any = dropout
a__ : int = attention_dropout
a__ : List[str] = gelu_activation
a__ : Optional[Any] = sinusoidal_embeddings
a__ : Optional[int] = causal
a__ : Optional[int] = asm
a__ : int = n_langs
a__ : Tuple = use_lang_emb
a__ : str = layer_norm_eps
a__ : Tuple = bos_index
a__ : int = eos_index
a__ : Optional[int] = pad_index
a__ : List[str] = unk_index
a__ : List[Any] = mask_index
a__ : List[str] = is_encoder
a__ : List[Any] = max_position_embeddings
a__ : Union[str, Any] = embed_init_std
a__ : Tuple = init_std
a__ : List[str] = summary_type
a__ : List[str] = summary_use_proj
a__ : Dict = summary_activation
a__ : str = summary_proj_to_labels
a__ : Dict = summary_first_dropout
a__ : List[str] = start_n_top
a__ : List[Any] = end_n_top
a__ : Dict = mask_token_id
a__ : Optional[Any] = lang_id
if "n_words" in kwargs:
a__ : Tuple = kwargs["n_words"]
super().__init__(pad_token_id=lowerCamelCase__ , bos_token_id=lowerCamelCase__ , **lowerCamelCase__ )
class A__ ( A__ ):
"""simple docstring"""
@property
def _UpperCamelCase( self : Dict ):
if self.task == "multiple-choice":
a__ : str = {0: "batch", 1: "choice", 2: "sequence"}
else:
a__ : Any = {0: "batch", 1: "sequence"}
return OrderedDict(
[
("input_ids", dynamic_axis),
("attention_mask", dynamic_axis),
("token_type_ids", dynamic_axis),
] )
| 37 |
import json
from typing import Dict, List, Optional, Tuple, Union
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import PaddingStrategy, logging
from .tokenization_led import LEDTokenizer
UpperCamelCase : Union[str, Any] = logging.get_logger(__name__)
UpperCamelCase : Optional[Any] = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Optional[Any] = {
"""vocab_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json""",
},
"""merges_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt""",
},
"""tokenizer_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json""",
},
}
UpperCamelCase : Dict = {
"""allenai/led-base-16384""": 1_6384,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = LEDTokenizer
_lowercase = ['input_ids', 'attention_mask']
def __init__( self : Tuple , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[str]=None , lowerCamelCase__ : Any=None , lowerCamelCase__ : int="replace" , lowerCamelCase__ : Union[str, Any]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Tuple="</s>" , lowerCamelCase__ : Optional[int]="<s>" , lowerCamelCase__ : str="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Any="<mask>" , lowerCamelCase__ : Optional[int]=False , lowerCamelCase__ : int=True , **lowerCamelCase__ : Union[str, Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : List[str] = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : Optional[Any] = add_prefix_space
a__ : List[str] = pre_tok_class(**lowerCamelCase__ )
a__ : Optional[int] = add_prefix_space
# the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__`
a__ : Any = "post_processor"
a__ : str = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Any = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Optional[Any] = tuple(state["sep"] )
if "cls" in state:
a__ : Optional[Any] = tuple(state["cls"] )
a__ : Optional[int] = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Dict = add_prefix_space
a__ : int = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : List[Any] = trim_offsets
a__ : List[str] = True
if changes_to_apply:
a__ : int = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : int = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
# Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Union[str, Any] ):
a__ : Any = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : Union[str, Any] = value
def _UpperCamelCase( self : Any , *lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : List[str] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
if is_split_into_words and not self.add_prefix_space:
raise ValueError(
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs." )
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Any , *lowerCamelCase__ : Dict , **lowerCamelCase__ : Optional[Any] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
if is_split_into_words and not self.add_prefix_space:
raise ValueError(
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs." )
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : List[str] = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[Any]=None ):
a__ : Any = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : List[str] = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Union[Dict[str, EncodedInput], BatchEncoding] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : PaddingStrategy = PaddingStrategy.DO_NOT_PAD , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[bool] = None , ):
a__ : str = super()._pad(
encoded_inputs=lowerCamelCase__ , max_length=lowerCamelCase__ , padding_strategy=lowerCamelCase__ , pad_to_multiple_of=lowerCamelCase__ , return_attention_mask=lowerCamelCase__ , )
# Load from model defaults
if return_attention_mask is None:
a__ : Optional[int] = "attention_mask" in self.model_input_names
if return_attention_mask and "global_attention_mask" in encoded_inputs:
a__ : Tuple = encoded_inputs[self.model_input_names[0]]
# `global_attention_mask` need to have the same length as other (sequential) inputs.
a__ : Dict = len(encoded_inputs["global_attention_mask"] ) != len(lowerCamelCase__ )
if needs_to_be_padded:
a__ : Union[str, Any] = len(lowerCamelCase__ ) - len(encoded_inputs["global_attention_mask"] )
if self.padding_side == "right":
# Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend`
a__ : List[Any] = (
encoded_inputs["global_attention_mask"] + [-1] * difference
)
elif self.padding_side == "left":
a__ : Any = [-1] * difference + encoded_inputs[
"global_attention_mask"
]
else:
raise ValueError("Invalid padding strategy:" + str(self.padding_side ) )
return encoded_inputs
| 37 | 1 |
import unittest
from transformers import is_tf_available
from transformers.testing_utils import require_tf
if is_tf_available():
import tensorflow as tf
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from transformers import GradientAccumulator, create_optimizer
@require_tf
class A__ ( unittest.TestCase ):
"""simple docstring"""
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : List[str] , lowerCamelCase__ : List[str] , lowerCamelCase__ : List[Any] ):
self.assertEqual(len(lowerCamelCase__ ) , len(lowerCamelCase__ ) )
for a, b in zip(lowerCamelCase__ , lowerCamelCase__ ):
self.assertAlmostEqual(lowerCamelCase__ , lowerCamelCase__ , delta=lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : List[str] = GradientAccumulator()
accumulator([tf.constant([1.0, 2.0] )] )
accumulator([tf.constant([-2.0, 1.0] )] )
accumulator([tf.constant([-1.0, 2.0] )] )
with self.assertRaises(lowerCamelCase__ ):
accumulator([tf.constant([1.0, 1.0] ), tf.constant([2.0, 2.0] )] )
self.assertEqual(accumulator.step , 3 )
self.assertEqual(len(accumulator.gradients ) , 1 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [-2.0, 5.0] , tol=1E-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
self.assertListAlmostEqual(accumulator.gradients[0].numpy().tolist() , [0.0, 0.0] , tol=1E-2 )
def _UpperCamelCase( self : List[Any] ):
a__ : Tuple = None
ops.enable_eager_execution_internal()
a__ : Dict = tf.config.list_physical_devices("CPU" )
if len(lowerCamelCase__ ) == 1:
tf.config.set_logical_device_configuration(
physical_devices[0] , [tf.config.LogicalDeviceConfiguration(), tf.config.LogicalDeviceConfiguration()] )
a__ : Optional[int] = tf.config.list_logical_devices(device_type="CPU" )
a__ : Tuple = tf.distribute.MirroredStrategy(devices=devices[:2] )
with strategy.scope():
a__ : List[Any] = GradientAccumulator()
a__ : str = tf.Variable([4.0, 3.0] )
a__, a__ : Union[str, Any] = create_optimizer(5E-5 , 10 , 5 )
a__ : Optional[int] = tf.Variable([0.0, 0.0] , trainable=lowerCamelCase__ )
def accumulate_on_replica(lowerCamelCase__ : Any ):
accumulator([gradient] )
def apply_on_replica():
optimizer.apply_gradients(list(zip(accumulator.gradients , [variable] ) ) )
@tf.function
def accumulate(lowerCamelCase__ : Any , lowerCamelCase__ : Optional[Any] ):
with strategy.scope():
a__ : Tuple = strategy.experimental_local_results(lowerCamelCase__ )
local_variables[0].assign(lowerCamelCase__ )
local_variables[1].assign(lowerCamelCase__ )
strategy.run(lowerCamelCase__ , args=(gradient_placeholder,) )
@tf.function
def apply_grad():
with strategy.scope():
strategy.run(lowerCamelCase__ )
def _check_local_values(lowerCamelCase__ : Tuple , lowerCamelCase__ : Optional[int] ):
a__ : Optional[int] = strategy.experimental_local_results(accumulator._gradients[0] )
self.assertListAlmostEqual(values[0].value() , lowerCamelCase__ , tol=1E-2 )
self.assertListAlmostEqual(values[1].value() , lowerCamelCase__ , tol=1E-2 )
accumulate([1.0, 2.0] , [-1.0, 1.0] )
accumulate([3.0, -1.0] , [-1.0, -1.0] )
accumulate([-2.0, 2.0] , [3.0, -2.0] )
self.assertEqual(accumulator.step , 3 )
_check_local_values([2.0, 3.0] , [1.0, -2.0] )
apply_grad()
self.assertListAlmostEqual(variable.value() , [4.0, 3.0] , tol=1E-2 )
accumulator.reset()
self.assertEqual(accumulator.step , 0 )
_check_local_values([0.0, 0.0] , [0.0, 0.0] )
| 37 |
import json
from typing import List, Optional, Tuple
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import logging
from .tokenization_roberta import RobertaTokenizer
UpperCamelCase : Any = logging.get_logger(__name__)
UpperCamelCase : Any = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Union[str, Any] = {
"""vocab_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/vocab.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/vocab.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/vocab.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/vocab.json""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/vocab.json""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/vocab.json"""
),
},
"""merges_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/merges.txt""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/merges.txt""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/merges.txt""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/merges.txt""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/merges.txt""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/merges.txt"""
),
},
"""tokenizer_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/tokenizer.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/tokenizer.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/tokenizer.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/tokenizer.json""",
"""roberta-base-openai-detector""": (
"""https://huggingface.co/roberta-base-openai-detector/resolve/main/tokenizer.json"""
),
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/tokenizer.json"""
),
},
}
UpperCamelCase : List[str] = {
"""roberta-base""": 512,
"""roberta-large""": 512,
"""roberta-large-mnli""": 512,
"""distilroberta-base""": 512,
"""roberta-base-openai-detector""": 512,
"""roberta-large-openai-detector""": 512,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = ['input_ids', 'attention_mask']
_lowercase = RobertaTokenizer
def __init__( self : List[str] , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[Any]=None , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]="replace" , lowerCamelCase__ : List[str]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Any="</s>" , lowerCamelCase__ : Any="<s>" , lowerCamelCase__ : int="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Tuple="<mask>" , lowerCamelCase__ : Any=False , lowerCamelCase__ : Dict=True , **lowerCamelCase__ : Optional[Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Any = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : int = add_prefix_space
a__ : Tuple = pre_tok_class(**lowerCamelCase__ )
a__ : str = add_prefix_space
a__ : Tuple = "post_processor"
a__ : Dict = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Tuple = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Tuple = tuple(state["sep"] )
if "cls" in state:
a__ : str = tuple(state["cls"] )
a__ : str = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : str = add_prefix_space
a__ : Any = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : int = trim_offsets
a__ : Dict = True
if changes_to_apply:
a__ : Union[str, Any] = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : str = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Tuple ):
a__ : List[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : List[str] = value
def _UpperCamelCase( self : Union[str, Any] , *lowerCamelCase__ : int , **lowerCamelCase__ : int ):
a__ : Optional[int] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Tuple , *lowerCamelCase__ : Dict , **lowerCamelCase__ : List[str] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : str , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : int = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[int]=None ):
a__ : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : Tuple = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
| 37 | 1 |
UpperCamelCase : Optional[int] = tuple[float, float, float]
UpperCamelCase : str = tuple[float, float, float]
def UpperCamelCase_ ( __a , __a ) -> Vectorad:
a__ : List[str] = end_pointa[0] - end_pointa[0]
a__ : List[Any] = end_pointa[1] - end_pointa[1]
a__ : Any = end_pointa[2] - end_pointa[2]
return (x, y, z)
def UpperCamelCase_ ( __a , __a ) -> Vectorad:
a__ : Dict = ab[1] * ac[2] - ab[2] * ac[1] # *i
a__ : int = (ab[0] * ac[2] - ab[2] * ac[0]) * -1 # *j
a__ : Any = ab[0] * ac[1] - ab[1] * ac[0] # *k
return (x, y, z)
def UpperCamelCase_ ( __a , __a ) -> bool:
return tuple(round(__a , __a ) for x in vector ) == (0, 0, 0)
def UpperCamelCase_ ( __a , __a , __a , __a = 10 ) -> bool:
a__ : Dict = create_vector(__a , __a )
a__ : Union[str, Any] = create_vector(__a , __a )
return is_zero_vector(get_ad_vectors_cross(__a , __a ) , __a )
| 37 |
from statistics import mean, stdev
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : List[str] = min(__a )
a__ : str = max(__a )
# normalize data
return [round((x - x_min) / (x_max - x_min) , __a ) for x in data]
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : str = mean(__a )
a__ : List[str] = stdev(__a )
# standardize data
return [round((x - mu) / (sigma) , __a ) for x in data]
| 37 | 1 |
def UpperCamelCase_ ( __a ) -> list[int]:
if length <= 0 or not isinstance(__a , __a ):
raise ValueError("Length must be a positive integer." )
return [n * (2 * n - 1) for n in range(__a )]
if __name__ == "__main__":
print(hexagonal_numbers(length=5))
print(hexagonal_numbers(length=10))
| 37 |
def UpperCamelCase_ ( __a = 50 ) -> int:
a__ : Tuple = [[0] * 3 for _ in range(length + 1 )]
for row_length in range(length + 1 ):
for tile_length in range(2 , 5 ):
for tile_start in range(row_length - tile_length + 1 ):
different_colour_ways_number[row_length][tile_length - 2] += (
different_colour_ways_number[row_length - tile_start - tile_length][
tile_length - 2
]
+ 1
)
return sum(different_colour_ways_number[length] )
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 | 1 |
import warnings
from ...utils import is_sklearn_available, requires_backends
if is_sklearn_available():
from scipy.stats import pearsonr, spearmanr
from sklearn.metrics import fa_score, matthews_corrcoef
UpperCamelCase : int = (
"""This metric will be removed from the library soon, metrics should be handled with the 🤗 Evaluate """
"""library. You can have a look at this example script for pointers: """
"""https://github.com/huggingface/transformers/blob/main/examples/pytorch/text-classification/run_glue.py"""
)
def UpperCamelCase_ ( __a , __a ) -> str:
warnings.warn(__a , __a )
requires_backends(__a , "sklearn" )
return (preds == labels).mean()
def UpperCamelCase_ ( __a , __a ) -> List[Any]:
warnings.warn(__a , __a )
requires_backends(__a , "sklearn" )
a__ : Optional[Any] = simple_accuracy(__a , __a )
a__ : Any = fa_score(y_true=__a , y_pred=__a )
return {
"acc": acc,
"f1": fa,
"acc_and_f1": (acc + fa) / 2,
}
def UpperCamelCase_ ( __a , __a ) -> str:
warnings.warn(__a , __a )
requires_backends(__a , "sklearn" )
a__ : Union[str, Any] = pearsonr(__a , __a )[0]
a__ : int = spearmanr(__a , __a )[0]
return {
"pearson": pearson_corr,
"spearmanr": spearman_corr,
"corr": (pearson_corr + spearman_corr) / 2,
}
def UpperCamelCase_ ( __a , __a , __a ) -> Tuple:
warnings.warn(__a , __a )
requires_backends(__a , "sklearn" )
assert len(__a ) == len(__a ), f'''Predictions and labels have mismatched lengths {len(__a )} and {len(__a )}'''
if task_name == "cola":
return {"mcc": matthews_corrcoef(__a , __a )}
elif task_name == "sst-2":
return {"acc": simple_accuracy(__a , __a )}
elif task_name == "mrpc":
return acc_and_fa(__a , __a )
elif task_name == "sts-b":
return pearson_and_spearman(__a , __a )
elif task_name == "qqp":
return acc_and_fa(__a , __a )
elif task_name == "mnli":
return {"mnli/acc": simple_accuracy(__a , __a )}
elif task_name == "mnli-mm":
return {"mnli-mm/acc": simple_accuracy(__a , __a )}
elif task_name == "qnli":
return {"acc": simple_accuracy(__a , __a )}
elif task_name == "rte":
return {"acc": simple_accuracy(__a , __a )}
elif task_name == "wnli":
return {"acc": simple_accuracy(__a , __a )}
elif task_name == "hans":
return {"acc": simple_accuracy(__a , __a )}
else:
raise KeyError(__a )
def UpperCamelCase_ ( __a , __a , __a ) -> Tuple:
warnings.warn(__a , __a )
requires_backends(__a , "sklearn" )
if len(__a ) != len(__a ):
raise ValueError(f'''Predictions and labels have mismatched lengths {len(__a )} and {len(__a )}''' )
if task_name == "xnli":
return {"acc": simple_accuracy(__a , __a )}
else:
raise KeyError(__a )
| 37 |
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[str] ):
a__ : str = name
a__ : Optional[int] = value
a__ : Dict = weight
def __repr__( self : Union[str, Any] ):
return f'''{self.__class__.__name__}({self.name}, {self.value}, {self.weight})'''
def _UpperCamelCase( self : Dict ):
return self.value
def _UpperCamelCase( self : Optional[Any] ):
return self.name
def _UpperCamelCase( self : Optional[Any] ):
return self.weight
def _UpperCamelCase( self : Optional[int] ):
return self.value / self.weight
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = []
for i in range(len(__a ) ):
menu.append(Things(name[i] , value[i] , weight[i] ) )
return menu
def UpperCamelCase_ ( __a , __a , __a ) -> Union[str, Any]:
a__ : List[str] = sorted(__a , key=__a , reverse=__a )
a__ : List[Any] = []
a__, a__ : Union[str, Any] = 0.0, 0.0
for i in range(len(__a ) ):
if (total_cost + items_copy[i].get_weight()) <= max_cost:
result.append(items_copy[i] )
total_cost += items_copy[i].get_weight()
total_value += items_copy[i].get_value()
return (result, total_value)
def UpperCamelCase_ ( ) -> Union[str, Any]:
pass
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 | 1 |
import importlib
import shutil
import threading
import warnings
from typing import List
import fsspec
import fsspec.asyn
from . import compression
from .hffilesystem import HfFileSystem
UpperCamelCase : List[Any] = importlib.util.find_spec("""s3fs""") is not None
if _has_safs:
from .safilesystem import SaFileSystem # noqa: F401
UpperCamelCase : List[compression.BaseCompressedFileFileSystem] = [
compression.BzaFileSystem,
compression.GzipFileSystem,
compression.LzaFileSystem,
compression.XzFileSystem,
compression.ZstdFileSystem,
]
# Register custom filesystems
for fs_class in COMPRESSION_FILESYSTEMS + [HfFileSystem]:
if fs_class.protocol in fsspec.registry and fsspec.registry[fs_class.protocol] is not fs_class:
warnings.warn(f"""A filesystem protocol was already set for {fs_class.protocol} and will be overwritten.""")
fsspec.register_implementation(fs_class.protocol, fs_class, clobber=True)
def UpperCamelCase_ ( __a ) -> str:
if "://" in dataset_path:
a__ : Optional[int] = dataset_path.split("://" )[1]
return dataset_path
def UpperCamelCase_ ( __a ) -> bool:
if fs is not None and fs.protocol != "file":
return True
else:
return False
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = not is_remote_filesystem(__a )
if is_local:
# LocalFileSystem.mv does copy + rm, it is more efficient to simply move a local directory
shutil.move(fs._strip_protocol(__a ) , fs._strip_protocol(__a ) )
else:
fs.mv(__a , __a , recursive=__a )
def UpperCamelCase_ ( ) -> None:
if hasattr(fsspec.asyn , "reset_lock" ):
# for future fsspec>2022.05.0
fsspec.asyn.reset_lock()
else:
a__ : List[str] = None
a__ : Any = None
a__ : List[Any] = threading.Lock()
| 37 |
import multiprocessing
from typing import TYPE_CHECKING, Optional, Union
from .. import Dataset, Features, config
from ..formatting import query_table
from ..packaged_modules.sql.sql import Sql
from ..utils import logging
from .abc import AbstractDatasetInputStream
if TYPE_CHECKING:
import sqlitea
import sqlalchemy
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Dict , lowerCamelCase__ : Union[str, "sqlalchemy.sql.Selectable"] , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[Features] = None , lowerCamelCase__ : str = None , lowerCamelCase__ : bool = False , **lowerCamelCase__ : Optional[int] , ):
super().__init__(features=lowerCamelCase__ , cache_dir=lowerCamelCase__ , keep_in_memory=lowerCamelCase__ , **lowerCamelCase__ )
a__ : str = Sql(
cache_dir=lowerCamelCase__ , features=lowerCamelCase__ , sql=lowerCamelCase__ , con=lowerCamelCase__ , **lowerCamelCase__ , )
def _UpperCamelCase( self : Tuple ):
a__ : Optional[Any] = None
a__ : Dict = None
a__ : Union[str, Any] = None
a__ : Union[str, Any] = None
self.builder.download_and_prepare(
download_config=lowerCamelCase__ , download_mode=lowerCamelCase__ , verification_mode=lowerCamelCase__ , base_path=lowerCamelCase__ , )
# Build dataset for splits
a__ : List[str] = self.builder.as_dataset(
split="train" , verification_mode=lowerCamelCase__ , in_memory=self.keep_in_memory )
return dataset
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : Dataset , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[int] = None , **lowerCamelCase__ : Optional[Any] , ):
if num_proc is not None and num_proc <= 0:
raise ValueError(f'''num_proc {num_proc} must be an integer > 0.''' )
a__ : Any = dataset
a__ : str = name
a__ : Tuple = con
a__ : List[Any] = batch_size if batch_size else config.DEFAULT_MAX_BATCH_SIZE
a__ : Any = num_proc
a__ : Tuple = to_sql_kwargs
def _UpperCamelCase( self : List[Any] ):
a__ : Any = self.to_sql_kwargs.pop("sql" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("con" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("index" , lowerCamelCase__ )
a__ : int = self._write(index=lowerCamelCase__ , **self.to_sql_kwargs )
return written
def _UpperCamelCase( self : Any , lowerCamelCase__ : List[str] ):
a__, a__, a__ : Union[str, Any] = args
a__ : Any = {**to_sql_kwargs, "if_exists": "append"} if offset > 0 else to_sql_kwargs
a__ : Tuple = query_table(
table=self.dataset.data , key=slice(lowerCamelCase__ , offset + self.batch_size ) , indices=self.dataset._indices , )
a__ : str = batch.to_pandas()
a__ : List[Any] = df.to_sql(self.name , self.con , index=lowerCamelCase__ , **lowerCamelCase__ )
return num_rows or len(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : str = 0
if self.num_proc is None or self.num_proc == 1:
for offset in logging.tqdm(
range(0 , len(self.dataset ) , self.batch_size ) , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += self._batch_sql((offset, index, to_sql_kwargs) )
else:
a__, a__ : List[str] = len(self.dataset ), self.batch_size
with multiprocessing.Pool(self.num_proc ) as pool:
for num_rows in logging.tqdm(
pool.imap(
self._batch_sql , [(offset, index, to_sql_kwargs) for offset in range(0 , lowerCamelCase__ , lowerCamelCase__ )] , ) , total=(num_rows // batch_size) + 1 if num_rows % batch_size else num_rows // batch_size , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += num_rows
return written
| 37 | 1 |
import unittest
from transformers import is_vision_available
from transformers.pipelines import pipeline
from transformers.testing_utils import (
is_pipeline_test,
nested_simplify,
require_tf,
require_torch,
require_vision,
slow,
)
from .test_pipelines_common import ANY
if is_vision_available():
from PIL import Image
else:
class A__ :
"""simple docstring"""
@staticmethod
def _UpperCamelCase( *lowerCamelCase__ : Dict , **lowerCamelCase__ : str ):
pass
@is_pipeline_test
@require_vision
class A__ ( unittest.TestCase ):
"""simple docstring"""
@require_torch
def _UpperCamelCase( self : Optional[int] ):
a__ : int = pipeline(
model="hf-internal-testing/tiny-random-clip-zero-shot-image-classification" , )
a__ : List[Any] = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
a__ : Optional[int] = image_classifier(lowerCamelCase__ , candidate_labels=["a", "b", "c"] )
# The floating scores are so close, we enter floating error approximation and the order is not guaranteed across
# python and torch versions.
self.assertIn(
nested_simplify(lowerCamelCase__ ) , [
[{"score": 0.333, "label": "a"}, {"score": 0.333, "label": "b"}, {"score": 0.333, "label": "c"}],
[{"score": 0.333, "label": "a"}, {"score": 0.333, "label": "c"}, {"score": 0.333, "label": "b"}],
] , )
a__ : Tuple = image_classifier([image] * 5 , candidate_labels=["A", "B", "C"] , batch_size=2 )
self.assertEqual(
nested_simplify(lowerCamelCase__ ) , [
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
] , )
@require_tf
def _UpperCamelCase( self : Optional[int] ):
a__ : List[str] = pipeline(
model="hf-internal-testing/tiny-random-clip-zero-shot-image-classification" , framework="tf" )
a__ : int = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
a__ : str = image_classifier(lowerCamelCase__ , candidate_labels=["a", "b", "c"] )
self.assertEqual(
nested_simplify(lowerCamelCase__ ) , [{"score": 0.333, "label": "a"}, {"score": 0.333, "label": "b"}, {"score": 0.333, "label": "c"}] , )
a__ : str = image_classifier([image] * 5 , candidate_labels=["A", "B", "C"] , batch_size=2 )
self.assertEqual(
nested_simplify(lowerCamelCase__ ) , [
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
[
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
{"score": 0.333, "label": ANY(lowerCamelCase__ )},
],
] , )
@slow
@require_torch
def _UpperCamelCase( self : str ):
a__ : List[Any] = pipeline(
task="zero-shot-image-classification" , model="openai/clip-vit-base-patch32" , )
# This is an image of 2 cats with remotes and no planes
a__ : Tuple = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
a__ : Any = image_classifier(lowerCamelCase__ , candidate_labels=["cat", "plane", "remote"] )
self.assertEqual(
nested_simplify(lowerCamelCase__ ) , [
{"score": 0.511, "label": "remote"},
{"score": 0.485, "label": "cat"},
{"score": 0.004, "label": "plane"},
] , )
a__ : int = image_classifier([image] * 5 , candidate_labels=["cat", "plane", "remote"] , batch_size=2 )
self.assertEqual(
nested_simplify(lowerCamelCase__ ) , [
[
{"score": 0.511, "label": "remote"},
{"score": 0.485, "label": "cat"},
{"score": 0.004, "label": "plane"},
],
]
* 5 , )
@slow
@require_tf
def _UpperCamelCase( self : Optional[int] ):
a__ : Dict = pipeline(
task="zero-shot-image-classification" , model="openai/clip-vit-base-patch32" , framework="tf" )
# This is an image of 2 cats with remotes and no planes
a__ : Optional[Any] = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
a__ : int = image_classifier(lowerCamelCase__ , candidate_labels=["cat", "plane", "remote"] )
self.assertEqual(
nested_simplify(lowerCamelCase__ ) , [
{"score": 0.511, "label": "remote"},
{"score": 0.485, "label": "cat"},
{"score": 0.004, "label": "plane"},
] , )
a__ : List[Any] = image_classifier([image] * 5 , candidate_labels=["cat", "plane", "remote"] , batch_size=2 )
self.assertEqual(
nested_simplify(lowerCamelCase__ ) , [
[
{"score": 0.511, "label": "remote"},
{"score": 0.485, "label": "cat"},
{"score": 0.004, "label": "plane"},
],
]
* 5 , )
| 37 |
import math
from datetime import datetime, timedelta
def UpperCamelCase_ ( __a ) -> datetime:
a__ : Union[str, Any] = year % 19
a__ : List[str] = year % 4
a__ : str = year % 7
a__ : Any = math.floor(year / 100 )
a__ : List[str] = math.floor((13 + 8 * leap_day_inhibits) / 25 )
a__ : Optional[int] = leap_day_inhibits / 4
a__ : Union[str, Any] = (
15 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 30
a__ : Dict = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
a__ : Any = (19 * metonic_cycle + secular_moon_shift) % 30
# PHM -> Paschal Full Moon
a__ : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 29 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 19 )
elif days_to_add == 28 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 18 )
else:
return datetime(__a , 3 , 22 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
UpperCamelCase : Tuple = """will be""" if year > datetime.now().year else """was"""
print(f"""Easter in {year} {tense} {gauss_easter(year)}""")
| 37 | 1 |
import gc
import importlib.metadata
import tempfile
import unittest
from packaging import version
from transformers import (
AutoModel,
AutoModelForCausalLM,
AutoModelForSeqaSeqLM,
AutoModelForSequenceClassification,
AutoTokenizer,
BitsAndBytesConfig,
pipeline,
)
from transformers.testing_utils import (
is_torch_available,
require_accelerate,
require_bitsandbytes,
require_torch,
require_torch_gpu,
require_torch_multi_gpu,
slow,
)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
if model.config.model_type == "gpt2":
return model.transformer.h[0].mlp.c_fc
return model.transformer.h[0].mlp.dense_ah_to_h
if is_torch_available():
import torch
import torch.nn as nn
class A__ ( nn.Module ):
"""simple docstring"""
def __init__( self : List[str] , lowerCamelCase__ : nn.Module , lowerCamelCase__ : int ):
super().__init__()
a__ : int = module
a__ : Any = nn.Sequential(
nn.Linear(module.in_features , lowerCamelCase__ , bias=lowerCamelCase__ ) , nn.Linear(lowerCamelCase__ , module.out_features , bias=lowerCamelCase__ ) , )
a__ : Tuple = (2.0 / (5 * min(module.in_features , module.out_features ))) ** 0.5
nn.init.normal_(self.adapter[0].weight , std=lowerCamelCase__ )
nn.init.zeros_(self.adapter[1].weight )
self.adapter.to(module.weight.device )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Optional[int] , *lowerCamelCase__ : int , **lowerCamelCase__ : Dict ):
return self.module(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ ) + self.adapter(lowerCamelCase__ )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
_lowercase = 'bigscience/bloom-1b7'
# Constant values
_lowercase = 2.1_09_65_95_52_69_25_74
_lowercase = 'Hello my name is'
_lowercase = set()
EXPECTED_OUTPUTS.add('Hello my name is John and I am a professional photographer. I' )
EXPECTED_OUTPUTS.add('Hello my name is John.\nI am a friend of your father.\n' )
EXPECTED_OUTPUTS.add('Hello my name is John Doe, I am a student at the University' )
_lowercase = 1_0
def _UpperCamelCase( self : Dict ):
# Models and tokenizer
a__ : List[str] = AutoTokenizer.from_pretrained(self.model_name )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Union[str, Any] ):
super().setUp()
# Models and tokenizer
a__ : List[Any] = AutoModelForCausalLM.from_pretrained(
self.model_name , torch_dtype=torch.floataa , device_map="auto" )
a__ : Union[str, Any] = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.model_fpaa
del self.model_abit
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : List[Any] ):
a__ : str = self.model_abit.config
self.assertTrue(hasattr(lowerCamelCase__ , "quantization_config" ) )
a__ : Optional[Any] = config.to_dict()
a__ : int = config.to_diff_dict()
a__ : List[str] = config.to_json_string()
def _UpperCamelCase( self : int ):
from bitsandbytes.nn import Paramsabit
a__ : List[Any] = self.model_fpaa.get_memory_footprint()
a__ : str = self.model_abit.get_memory_footprint()
self.assertAlmostEqual(mem_fpaa / mem_abit , self.EXPECTED_RELATIVE_DIFFERENCE )
a__ : Optional[Any] = get_some_linear_layer(self.model_abit )
self.assertTrue(linear.weight.__class__ == Paramsabit )
def _UpperCamelCase( self : Tuple ):
from transformers import TaPreTrainedModel
self.model_fpaa.get_memory_footprint()
self.model_abit.get_memory_footprint()
for name, module in self.model_abit.named_modules():
if isinstance(lowerCamelCase__ , torch.nn.Linear ):
if name not in ["lm_head"] + TaPreTrainedModel._keep_in_fpaa_modules:
# 4-bit parameters are packed in uint8 variables
self.assertTrue(module.weight.dtype == torch.uinta )
def _UpperCamelCase( self : str ):
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Tuple = self.model_abit.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[Any] = BitsAndBytesConfig()
a__ : Optional[int] = True
a__ : int = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , device_map="auto" )
a__ : str = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : int = model_abit_from_config.generate(
input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : Dict ):
with self.assertRaises(lowerCamelCase__ ), tempfile.TemporaryDirectory() as tmpdirname:
self.model_abit.save_pretrained(lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : int = BitsAndBytesConfig()
with self.assertRaises(lowerCamelCase__ ):
a__ : Dict = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , load_in_abit=lowerCamelCase__ , device_map="auto" , bnb_abit_quant_type="nf4" , )
def _UpperCamelCase( self : int ):
with self.assertRaises(lowerCamelCase__ ):
# Tries with `str`
self.model_abit.to("cpu" )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `dtype``
self.model_abit.to(torch.floataa )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.to(torch.device("cuda:0" ) )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.float()
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.half()
# Test if we did not break anything
a__ : int = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Any = self.model_fpaa.to(torch.floataa )
a__ : List[Any] = self.model_fpaa.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.to("cpu" )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.half()
# Check this does not throw an error
a__ : Dict = self.model_fpaa.float()
def _UpperCamelCase( self : Dict ):
a__ : List[str] = AutoModelForSeqaSeqLM.from_pretrained("t5-small" , load_in_abit=lowerCamelCase__ , device_map="auto" )
self.assertTrue(model.decoder.block[0].layer[2].DenseReluDense.wo.weight.dtype == torch.floataa )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
@classmethod
def _UpperCamelCase( cls : str ):
a__ : Dict = "t5-small"
a__ : List[Any] = "google/flan-t5-small" # flan-t5 uses dense-act instead of dense-relu-dense
a__ : int = AutoTokenizer.from_pretrained(cls.model_name )
a__ : str = "Translate in German: Hello, my dog is cute"
def _UpperCamelCase( self : Optional[int] ):
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Optional[int] ):
from transformers import TaForConditionalGeneration
a__ : List[Any] = TaForConditionalGeneration._keep_in_fpaa_modules
a__ : Optional[Any] = None
# test with `t5-small`
a__ : Any = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : Dict = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Any = model.generate(**lowerCamelCase__ )
a__ : Union[str, Any] = modules
def _UpperCamelCase( self : List[Any] ):
import bitsandbytes as bnb
from transformers import TaForConditionalGeneration
# test with `t5-small`
a__ : List[str] = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# there was a bug with decoders - this test checks that it is fixed
self.assertTrue(isinstance(model.decoder.block[0].layer[0].SelfAttention.q , bnb.nn.Linearabit ) )
a__ : Union[str, Any] = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : int = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Any = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Optional[int] = model.generate(**lowerCamelCase__ )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : List[str] ):
super().setUp()
# model_name
a__ : Union[str, Any] = "bigscience/bloom-560m"
a__ : Union[str, Any] = "t5-small"
# Different types of model
a__ : int = AutoModel.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Sequence classification model
a__ : Dict = AutoModelForSequenceClassification.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# CausalLM model
a__ : str = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Seq2seq model
a__ : Dict = AutoModelForSeqaSeqLM.from_pretrained(
self.seq_to_seq_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.base_model
del self.sequence_model
del self.model_abit
del self.seq_to_seq_model
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Union[str, Any] ):
from bitsandbytes.nn import Paramsabit
self.assertTrue(self.base_model.h[-1].mlp.dense_ah_to_h.weight.__class__ == Paramsabit )
# Other heads should be nn.Parameter
self.assertTrue(self.model_abit.lm_head.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.sequence_model.score.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.seq_to_seq_model.lm_head.weight.__class__ == torch.nn.Parameter )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
super().setUp()
def _UpperCamelCase( self : int ):
del self.pipe
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Tuple ):
a__ : int = pipeline(
"text-generation" , model=self.model_name , model_kwargs={"device_map": "auto", "load_in_4bit": True, "torch_dtype": torch.floataa} , max_new_tokens=self.MAX_NEW_TOKENS , )
# Real second forward pass
a__ : Tuple = self.pipe(self.input_text )
self.assertIn(pipeline_output[0]["generated_text"] , self.EXPECTED_OUTPUTS )
@require_torch_multi_gpu
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Tuple ):
super().setUp()
def _UpperCamelCase( self : List[Any] ):
a__ : str = AutoModelForCausalLM.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="balanced" )
# Check correct device map
self.assertEqual(set(model_parallel.hf_device_map.values() ) , {0, 1} )
# Check that inference pass works on the model
a__ : List[Any] = self.tokenizer(self.input_text , return_tensors="pt" )
# Second real batch
a__ : List[Any] = model_parallel.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_parallel[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
a__ : Any = "facebook/opt-350m"
super().setUp()
def _UpperCamelCase( self : int ):
if version.parse(importlib.metadata.version("bitsandbytes" ) ) < version.parse("0.37.0" ):
return
# Step 1: freeze all parameters
a__ : Tuple = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ )
self.assertEqual(set(model.hf_device_map.values() ) , {torch.cuda.current_device()} )
for param in model.parameters():
a__ : Any = False # freeze the model - train adapters later
if param.ndim == 1:
# cast the small parameters (e.g. layernorm) to fp32 for stability
a__ : Tuple = param.data.to(torch.floataa )
# Step 2: add adapters
for _, module in model.named_modules():
if "OPTAttention" in repr(type(lowerCamelCase__ ) ):
a__ : Dict = LoRALayer(module.q_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.k_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.v_proj , rank=16 )
# Step 3: dummy batch
a__ : Dict = self.tokenizer("Test batch " , return_tensors="pt" ).to(0 )
# Step 4: Check if the gradient is not None
with torch.cuda.amp.autocast():
a__ : Optional[Any] = model.forward(**lowerCamelCase__ )
out.logits.norm().backward()
for module in model.modules():
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertTrue(module.adapter[1].weight.grad is not None )
self.assertTrue(module.adapter[1].weight.grad.norm().item() > 0 )
elif isinstance(lowerCamelCase__ , nn.Embedding ):
self.assertTrue(module.weight.grad is None )
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gpt2-xl'
_lowercase = 3.31_91_85_48_54_15_21_87
| 37 |
import gc
import importlib.metadata
import tempfile
import unittest
from packaging import version
from transformers import (
AutoModel,
AutoModelForCausalLM,
AutoModelForSeqaSeqLM,
AutoModelForSequenceClassification,
AutoTokenizer,
BitsAndBytesConfig,
pipeline,
)
from transformers.testing_utils import (
is_torch_available,
require_accelerate,
require_bitsandbytes,
require_torch,
require_torch_gpu,
require_torch_multi_gpu,
slow,
)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
if model.config.model_type == "gpt2":
return model.transformer.h[0].mlp.c_fc
return model.transformer.h[0].mlp.dense_ah_to_h
if is_torch_available():
import torch
import torch.nn as nn
class A__ ( nn.Module ):
"""simple docstring"""
def __init__( self : List[str] , lowerCamelCase__ : nn.Module , lowerCamelCase__ : int ):
super().__init__()
a__ : int = module
a__ : Any = nn.Sequential(
nn.Linear(module.in_features , lowerCamelCase__ , bias=lowerCamelCase__ ) , nn.Linear(lowerCamelCase__ , module.out_features , bias=lowerCamelCase__ ) , )
a__ : Tuple = (2.0 / (5 * min(module.in_features , module.out_features ))) ** 0.5
nn.init.normal_(self.adapter[0].weight , std=lowerCamelCase__ )
nn.init.zeros_(self.adapter[1].weight )
self.adapter.to(module.weight.device )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Optional[int] , *lowerCamelCase__ : int , **lowerCamelCase__ : Dict ):
return self.module(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ ) + self.adapter(lowerCamelCase__ )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
_lowercase = 'bigscience/bloom-1b7'
# Constant values
_lowercase = 2.1_09_65_95_52_69_25_74
_lowercase = 'Hello my name is'
_lowercase = set()
EXPECTED_OUTPUTS.add('Hello my name is John and I am a professional photographer. I' )
EXPECTED_OUTPUTS.add('Hello my name is John.\nI am a friend of your father.\n' )
EXPECTED_OUTPUTS.add('Hello my name is John Doe, I am a student at the University' )
_lowercase = 1_0
def _UpperCamelCase( self : Dict ):
# Models and tokenizer
a__ : List[str] = AutoTokenizer.from_pretrained(self.model_name )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Union[str, Any] ):
super().setUp()
# Models and tokenizer
a__ : List[Any] = AutoModelForCausalLM.from_pretrained(
self.model_name , torch_dtype=torch.floataa , device_map="auto" )
a__ : Union[str, Any] = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.model_fpaa
del self.model_abit
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : List[Any] ):
a__ : str = self.model_abit.config
self.assertTrue(hasattr(lowerCamelCase__ , "quantization_config" ) )
a__ : Optional[Any] = config.to_dict()
a__ : int = config.to_diff_dict()
a__ : List[str] = config.to_json_string()
def _UpperCamelCase( self : int ):
from bitsandbytes.nn import Paramsabit
a__ : List[Any] = self.model_fpaa.get_memory_footprint()
a__ : str = self.model_abit.get_memory_footprint()
self.assertAlmostEqual(mem_fpaa / mem_abit , self.EXPECTED_RELATIVE_DIFFERENCE )
a__ : Optional[Any] = get_some_linear_layer(self.model_abit )
self.assertTrue(linear.weight.__class__ == Paramsabit )
def _UpperCamelCase( self : Tuple ):
from transformers import TaPreTrainedModel
self.model_fpaa.get_memory_footprint()
self.model_abit.get_memory_footprint()
for name, module in self.model_abit.named_modules():
if isinstance(lowerCamelCase__ , torch.nn.Linear ):
if name not in ["lm_head"] + TaPreTrainedModel._keep_in_fpaa_modules:
# 4-bit parameters are packed in uint8 variables
self.assertTrue(module.weight.dtype == torch.uinta )
def _UpperCamelCase( self : str ):
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Tuple = self.model_abit.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[Any] = BitsAndBytesConfig()
a__ : Optional[int] = True
a__ : int = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , device_map="auto" )
a__ : str = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : int = model_abit_from_config.generate(
input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : Dict ):
with self.assertRaises(lowerCamelCase__ ), tempfile.TemporaryDirectory() as tmpdirname:
self.model_abit.save_pretrained(lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : int = BitsAndBytesConfig()
with self.assertRaises(lowerCamelCase__ ):
a__ : Dict = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , load_in_abit=lowerCamelCase__ , device_map="auto" , bnb_abit_quant_type="nf4" , )
def _UpperCamelCase( self : int ):
with self.assertRaises(lowerCamelCase__ ):
# Tries with `str`
self.model_abit.to("cpu" )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `dtype``
self.model_abit.to(torch.floataa )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.to(torch.device("cuda:0" ) )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.float()
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.half()
# Test if we did not break anything
a__ : int = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Any = self.model_fpaa.to(torch.floataa )
a__ : List[Any] = self.model_fpaa.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.to("cpu" )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.half()
# Check this does not throw an error
a__ : Dict = self.model_fpaa.float()
def _UpperCamelCase( self : Dict ):
a__ : List[str] = AutoModelForSeqaSeqLM.from_pretrained("t5-small" , load_in_abit=lowerCamelCase__ , device_map="auto" )
self.assertTrue(model.decoder.block[0].layer[2].DenseReluDense.wo.weight.dtype == torch.floataa )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
@classmethod
def _UpperCamelCase( cls : str ):
a__ : Dict = "t5-small"
a__ : List[Any] = "google/flan-t5-small" # flan-t5 uses dense-act instead of dense-relu-dense
a__ : int = AutoTokenizer.from_pretrained(cls.model_name )
a__ : str = "Translate in German: Hello, my dog is cute"
def _UpperCamelCase( self : Optional[int] ):
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Optional[int] ):
from transformers import TaForConditionalGeneration
a__ : List[Any] = TaForConditionalGeneration._keep_in_fpaa_modules
a__ : Optional[Any] = None
# test with `t5-small`
a__ : Any = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : Dict = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Any = model.generate(**lowerCamelCase__ )
a__ : Union[str, Any] = modules
def _UpperCamelCase( self : List[Any] ):
import bitsandbytes as bnb
from transformers import TaForConditionalGeneration
# test with `t5-small`
a__ : List[str] = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# there was a bug with decoders - this test checks that it is fixed
self.assertTrue(isinstance(model.decoder.block[0].layer[0].SelfAttention.q , bnb.nn.Linearabit ) )
a__ : Union[str, Any] = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : int = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Any = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Optional[int] = model.generate(**lowerCamelCase__ )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : List[str] ):
super().setUp()
# model_name
a__ : Union[str, Any] = "bigscience/bloom-560m"
a__ : Union[str, Any] = "t5-small"
# Different types of model
a__ : int = AutoModel.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Sequence classification model
a__ : Dict = AutoModelForSequenceClassification.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# CausalLM model
a__ : str = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Seq2seq model
a__ : Dict = AutoModelForSeqaSeqLM.from_pretrained(
self.seq_to_seq_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.base_model
del self.sequence_model
del self.model_abit
del self.seq_to_seq_model
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Union[str, Any] ):
from bitsandbytes.nn import Paramsabit
self.assertTrue(self.base_model.h[-1].mlp.dense_ah_to_h.weight.__class__ == Paramsabit )
# Other heads should be nn.Parameter
self.assertTrue(self.model_abit.lm_head.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.sequence_model.score.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.seq_to_seq_model.lm_head.weight.__class__ == torch.nn.Parameter )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
super().setUp()
def _UpperCamelCase( self : int ):
del self.pipe
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Tuple ):
a__ : int = pipeline(
"text-generation" , model=self.model_name , model_kwargs={"device_map": "auto", "load_in_4bit": True, "torch_dtype": torch.floataa} , max_new_tokens=self.MAX_NEW_TOKENS , )
# Real second forward pass
a__ : Tuple = self.pipe(self.input_text )
self.assertIn(pipeline_output[0]["generated_text"] , self.EXPECTED_OUTPUTS )
@require_torch_multi_gpu
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Tuple ):
super().setUp()
def _UpperCamelCase( self : List[Any] ):
a__ : str = AutoModelForCausalLM.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="balanced" )
# Check correct device map
self.assertEqual(set(model_parallel.hf_device_map.values() ) , {0, 1} )
# Check that inference pass works on the model
a__ : List[Any] = self.tokenizer(self.input_text , return_tensors="pt" )
# Second real batch
a__ : List[Any] = model_parallel.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_parallel[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
a__ : Any = "facebook/opt-350m"
super().setUp()
def _UpperCamelCase( self : int ):
if version.parse(importlib.metadata.version("bitsandbytes" ) ) < version.parse("0.37.0" ):
return
# Step 1: freeze all parameters
a__ : Tuple = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ )
self.assertEqual(set(model.hf_device_map.values() ) , {torch.cuda.current_device()} )
for param in model.parameters():
a__ : Any = False # freeze the model - train adapters later
if param.ndim == 1:
# cast the small parameters (e.g. layernorm) to fp32 for stability
a__ : Tuple = param.data.to(torch.floataa )
# Step 2: add adapters
for _, module in model.named_modules():
if "OPTAttention" in repr(type(lowerCamelCase__ ) ):
a__ : Dict = LoRALayer(module.q_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.k_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.v_proj , rank=16 )
# Step 3: dummy batch
a__ : Dict = self.tokenizer("Test batch " , return_tensors="pt" ).to(0 )
# Step 4: Check if the gradient is not None
with torch.cuda.amp.autocast():
a__ : Optional[Any] = model.forward(**lowerCamelCase__ )
out.logits.norm().backward()
for module in model.modules():
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertTrue(module.adapter[1].weight.grad is not None )
self.assertTrue(module.adapter[1].weight.grad.norm().item() > 0 )
elif isinstance(lowerCamelCase__ , nn.Embedding ):
self.assertTrue(module.weight.grad is None )
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gpt2-xl'
_lowercase = 3.31_91_85_48_54_15_21_87
| 37 | 1 |
from __future__ import annotations
from collections import namedtuple
def UpperCamelCase_ ( __a , __a , __a ) -> tuple:
a__ : Union[str, Any] = namedtuple("result" , "name value" )
if (voltage, current, power).count(0 ) != 1:
raise ValueError("Only one argument must be 0" )
elif power < 0:
raise ValueError(
"Power cannot be negative in any electrical/electronics system" )
elif voltage == 0:
return result("voltage" , power / current )
elif current == 0:
return result("current" , power / voltage )
elif power == 0:
return result("power" , float(round(abs(voltage * current ) , 2 ) ) )
else:
raise ValueError("Exactly one argument must be 0" )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 |
import inspect
import unittest
from datasets import load_dataset
from packaging import version
from transformers import BeitConfig
from transformers.models.auto import get_values
from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_vision_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, _config_zero_init, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import (
MODEL_MAPPING,
BeitForImageClassification,
BeitForMaskedImageModeling,
BeitForSemanticSegmentation,
BeitModel,
)
from transformers.models.beit.modeling_beit import BEIT_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
import PIL
from PIL import Image
from transformers import BeitImageProcessor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Optional[int]=100 , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[int]=30 , lowerCamelCase__ : Union[str, Any]=2 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : List[Any]=True , lowerCamelCase__ : Tuple=True , lowerCamelCase__ : int=32 , lowerCamelCase__ : Union[str, Any]=4 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Union[str, Any]=37 , lowerCamelCase__ : List[Any]="gelu" , lowerCamelCase__ : List[Any]=0.1 , lowerCamelCase__ : int=0.1 , lowerCamelCase__ : Union[str, Any]=10 , lowerCamelCase__ : str=0.02 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]=[0, 1, 2, 3] , ):
a__ : Dict = parent
a__ : Dict = 100
a__ : Optional[int] = batch_size
a__ : Union[str, Any] = image_size
a__ : Any = patch_size
a__ : Optional[Any] = num_channels
a__ : int = is_training
a__ : List[str] = use_labels
a__ : Optional[Any] = hidden_size
a__ : List[Any] = num_hidden_layers
a__ : str = num_attention_heads
a__ : str = intermediate_size
a__ : int = hidden_act
a__ : List[Any] = hidden_dropout_prob
a__ : Dict = attention_probs_dropout_prob
a__ : Union[str, Any] = type_sequence_label_size
a__ : Optional[Any] = initializer_range
a__ : List[str] = scope
a__ : int = out_indices
a__ : List[str] = num_labels
# in BeiT, the seq length equals the number of patches + 1 (we add 1 for the [CLS] token)
a__ : Optional[int] = (image_size // patch_size) ** 2
a__ : Union[str, Any] = num_patches + 1
def _UpperCamelCase( self : int ):
a__ : Optional[int] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
a__ : Optional[Any] = None
a__ : Tuple = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : Dict = ids_tensor([self.batch_size, self.image_size, self.image_size] , self.num_labels )
a__ : Optional[int] = self.get_config()
return config, pixel_values, labels, pixel_labels
def _UpperCamelCase( self : Tuple ):
return BeitConfig(
vocab_size=self.vocab_size , image_size=self.image_size , patch_size=self.patch_size , num_channels=self.num_channels , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , out_indices=self.out_indices , )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : int , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any ):
a__ : str = BeitModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[str] , lowerCamelCase__ : Any , lowerCamelCase__ : List[str] , lowerCamelCase__ : Tuple ):
a__ : int = BeitForMaskedImageModeling(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[Any] = model(lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length - 1, self.vocab_size) )
def _UpperCamelCase( self : str , lowerCamelCase__ : Any , lowerCamelCase__ : str , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Any ):
a__ : List[str] = self.type_sequence_label_size
a__ : Optional[Any] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
# test greyscale images
a__ : Optional[Any] = 1
a__ : List[str] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = floats_tensor([self.batch_size, 1, self.image_size, self.image_size] )
a__ : Union[str, Any] = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def _UpperCamelCase( self : Any , lowerCamelCase__ : str , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
a__ : int = self.num_labels
a__ : List[str] = BeitForSemanticSegmentation(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Tuple = model(lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
def _UpperCamelCase( self : Optional[int] ):
a__ : Any = self.prepare_config_and_inputs()
a__, a__, a__, a__ : Union[str, Any] = config_and_inputs
a__ : Dict = {"pixel_values": pixel_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(BeitModel, BeitForImageClassification, BeitForMaskedImageModeling, BeitForSemanticSegmentation)
if is_torch_available()
else ()
)
_lowercase = (
{
'feature-extraction': BeitModel,
'image-classification': BeitForImageClassification,
'image-segmentation': BeitForSemanticSegmentation,
}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Any ):
a__ : int = BeitModelTester(self )
a__ : Optional[Any] = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[Any] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="BEiT does not use inputs_embeds" )
def _UpperCamelCase( self : str ):
pass
@require_torch_multi_gpu
@unittest.skip(reason="BEiT has some layers using `add_module` which doesn't work well with `nn.DataParallel`" )
def _UpperCamelCase( self : Dict ):
pass
def _UpperCamelCase( self : Optional[Any] ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : List[str] = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Optional[int] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : str ):
a__, a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : int = model_class(lowerCamelCase__ )
a__ : str = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Any = ["pixel_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
def _UpperCamelCase( self : int ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_lm(*lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] ):
a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_semantic_segmentation(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
if not self.model_tester.is_training:
return
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : str = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]:
continue
a__ : List[str] = model_class(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.train()
a__ : Any = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : Tuple = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : Tuple ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
if not self.model_tester.is_training:
return
a__ : List[Any] = False
a__ : List[str] = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if (
model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]
or not model_class.supports_gradient_checkpointing
):
continue
a__ : Optional[Any] = model_class(lowerCamelCase__ )
model.gradient_checkpointing_enable()
model.to(lowerCamelCase__ )
model.train()
a__ : Union[str, Any] = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : int = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : List[str] ):
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : Dict = _config_zero_init(lowerCamelCase__ )
for model_class in self.all_model_classes:
a__ : str = model_class(config=lowerCamelCase__ )
for name, param in model.named_parameters():
# we skip lambda parameters as these require special initial values
# determined by config.layer_scale_init_value
if "lambda" in name:
continue
if param.requires_grad:
self.assertIn(
((param.data.mean() * 1E9).round() / 1E9).item() , [0.0, 1.0] , msg=f'''Parameter {name} of model {model_class} seems not properly initialized''' , )
@slow
def _UpperCamelCase( self : Optional[int] ):
for model_name in BEIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Tuple = BeitModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : List[Any] = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
return image
@require_torch
@require_vision
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : Optional[int] ):
return BeitImageProcessor.from_pretrained("microsoft/beit-base-patch16-224" ) if is_vision_available() else None
@slow
def _UpperCamelCase( self : str ):
a__ : int = BeitForMaskedImageModeling.from_pretrained("microsoft/beit-base-patch16-224-pt22k" ).to(lowerCamelCase__ )
a__ : Optional[Any] = self.default_image_processor
a__ : Dict = prepare_img()
a__ : Optional[int] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).pixel_values.to(lowerCamelCase__ )
# prepare bool_masked_pos
a__ : Optional[Any] = torch.ones((1, 196) , dtype=torch.bool ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(pixel_values=lowerCamelCase__ , bool_masked_pos=lowerCamelCase__ )
a__ : Tuple = outputs.logits
# verify the logits
a__ : List[str] = torch.Size((1, 196, 8_192) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[int] = torch.tensor(
[[-3.2437, 0.5072, -13.9174], [-3.2456, 0.4948, -13.9401], [-3.2033, 0.5121, -13.8550]] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[bool_masked_pos][:3, :3] , lowerCamelCase__ , atol=1E-2 ) )
@slow
def _UpperCamelCase( self : Dict ):
a__ : str = BeitForImageClassification.from_pretrained("microsoft/beit-base-patch16-224" ).to(lowerCamelCase__ )
a__ : int = self.default_image_processor
a__ : List[Any] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Union[str, Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 1_000) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = torch.tensor([-1.2385, -1.0987, -1.0108] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Tuple = 281
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : Any ):
a__ : Dict = BeitForImageClassification.from_pretrained("microsoft/beit-large-patch16-224-pt22k-ft22k" ).to(
lowerCamelCase__ )
a__ : str = self.default_image_processor
a__ : List[str] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Dict = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Optional[int] = torch.Size((1, 21_841) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[Any] = torch.tensor([1.6881, -0.2787, 0.5901] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Optional[Any] = 2_396
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
a__ : Optional[Any] = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : Tuple = model.to(lowerCamelCase__ )
a__ : List[Any] = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Tuple = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : Union[str, Any] = Image.open(ds[0]["file"] )
a__ : List[Any] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Optional[Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Tuple = torch.Size((1, 150, 160, 160) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = version.parse(PIL.__version__ ) < version.parse("9.0.0" )
if is_pillow_less_than_a:
a__ : Dict = torch.tensor(
[
[[-4.9225, -2.3954, -3.0522], [-2.8822, -1.0046, -1.7561], [-2.9549, -1.3228, -2.1347]],
[[-5.8168, -3.4129, -4.0778], [-3.8651, -2.2214, -3.0277], [-3.8356, -2.4643, -3.3535]],
[[-0.0078, 3.9952, 4.0754], [2.9856, 4.6944, 5.0035], [3.2413, 4.7813, 4.9969]],
] , device=lowerCamelCase__ , )
else:
a__ : Dict = torch.tensor(
[
[[-4.8960, -2.3688, -3.0355], [-2.8478, -0.9836, -1.7418], [-2.9449, -1.3332, -2.1456]],
[[-5.8081, -3.4124, -4.1006], [-3.8561, -2.2081, -3.0323], [-3.8365, -2.4601, -3.3669]],
[[-0.0309, 3.9868, 4.0540], [2.9640, 4.6877, 4.9976], [3.2081, 4.7690, 4.9942]],
] , device=lowerCamelCase__ , )
self.assertTrue(torch.allclose(logits[0, :3, :3, :3] , lowerCamelCase__ , atol=1E-4 ) )
@slow
def _UpperCamelCase( self : Tuple ):
a__ : str = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : List[Any] = model.to(lowerCamelCase__ )
a__ : int = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Optional[int] = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : str = Image.open(ds[0]["file"] )
a__ : str = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : List[Any] = model(**lowerCamelCase__ )
a__ : Any = outputs.logits.detach().cpu()
a__ : List[Any] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ , target_sizes=[(500, 300)] )
a__ : Optional[int] = torch.Size((500, 300) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
a__ : List[str] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ )
a__ : Any = torch.Size((160, 160) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
| 37 | 1 |
import json
from typing import List, Optional, Tuple
from tokenizers import normalizers
from ...tokenization_utils_base import BatchEncoding
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import PaddingStrategy, logging
from .tokenization_realm import RealmTokenizer
UpperCamelCase : List[str] = logging.get_logger(__name__)
UpperCamelCase : List[Any] = {"""vocab_file""": """vocab.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Union[str, Any] = {
"""vocab_file""": {
"""google/realm-cc-news-pretrained-embedder""": (
"""https://huggingface.co/google/realm-cc-news-pretrained-embedder/resolve/main/vocab.txt"""
),
"""google/realm-cc-news-pretrained-encoder""": (
"""https://huggingface.co/google/realm-cc-news-pretrained-encoder/resolve/main/vocab.txt"""
),
"""google/realm-cc-news-pretrained-scorer""": (
"""https://huggingface.co/google/realm-cc-news-pretrained-scorer/resolve/main/vocab.txt"""
),
"""google/realm-cc-news-pretrained-openqa""": (
"""https://huggingface.co/google/realm-cc-news-pretrained-openqa/aresolve/main/vocab.txt"""
),
"""google/realm-orqa-nq-openqa""": """https://huggingface.co/google/realm-orqa-nq-openqa/resolve/main/vocab.txt""",
"""google/realm-orqa-nq-reader""": """https://huggingface.co/google/realm-orqa-nq-reader/resolve/main/vocab.txt""",
"""google/realm-orqa-wq-openqa""": """https://huggingface.co/google/realm-orqa-wq-openqa/resolve/main/vocab.txt""",
"""google/realm-orqa-wq-reader""": """https://huggingface.co/google/realm-orqa-wq-reader/resolve/main/vocab.txt""",
},
"""tokenizer_file""": {
"""google/realm-cc-news-pretrained-embedder""": (
"""https://huggingface.co/google/realm-cc-news-pretrained-embedder/resolve/main/tokenizer.jsont"""
),
"""google/realm-cc-news-pretrained-encoder""": (
"""https://huggingface.co/google/realm-cc-news-pretrained-encoder/resolve/main/tokenizer.json"""
),
"""google/realm-cc-news-pretrained-scorer""": (
"""https://huggingface.co/google/realm-cc-news-pretrained-scorer/resolve/main/tokenizer.json"""
),
"""google/realm-cc-news-pretrained-openqa""": (
"""https://huggingface.co/google/realm-cc-news-pretrained-openqa/aresolve/main/tokenizer.json"""
),
"""google/realm-orqa-nq-openqa""": (
"""https://huggingface.co/google/realm-orqa-nq-openqa/resolve/main/tokenizer.json"""
),
"""google/realm-orqa-nq-reader""": (
"""https://huggingface.co/google/realm-orqa-nq-reader/resolve/main/tokenizer.json"""
),
"""google/realm-orqa-wq-openqa""": (
"""https://huggingface.co/google/realm-orqa-wq-openqa/resolve/main/tokenizer.json"""
),
"""google/realm-orqa-wq-reader""": (
"""https://huggingface.co/google/realm-orqa-wq-reader/resolve/main/tokenizer.json"""
),
},
}
UpperCamelCase : Optional[int] = {
"""google/realm-cc-news-pretrained-embedder""": 512,
"""google/realm-cc-news-pretrained-encoder""": 512,
"""google/realm-cc-news-pretrained-scorer""": 512,
"""google/realm-cc-news-pretrained-openqa""": 512,
"""google/realm-orqa-nq-openqa""": 512,
"""google/realm-orqa-nq-reader""": 512,
"""google/realm-orqa-wq-openqa""": 512,
"""google/realm-orqa-wq-reader""": 512,
}
UpperCamelCase : List[str] = {
"""google/realm-cc-news-pretrained-embedder""": {"""do_lower_case""": True},
"""google/realm-cc-news-pretrained-encoder""": {"""do_lower_case""": True},
"""google/realm-cc-news-pretrained-scorer""": {"""do_lower_case""": True},
"""google/realm-cc-news-pretrained-openqa""": {"""do_lower_case""": True},
"""google/realm-orqa-nq-openqa""": {"""do_lower_case""": True},
"""google/realm-orqa-nq-reader""": {"""do_lower_case""": True},
"""google/realm-orqa-wq-openqa""": {"""do_lower_case""": True},
"""google/realm-orqa-wq-reader""": {"""do_lower_case""": True},
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_INIT_CONFIGURATION
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = RealmTokenizer
def __init__( self : str , lowerCamelCase__ : str=None , lowerCamelCase__ : int=None , lowerCamelCase__ : Any=True , lowerCamelCase__ : Any="[UNK]" , lowerCamelCase__ : Optional[int]="[SEP]" , lowerCamelCase__ : Optional[Any]="[PAD]" , lowerCamelCase__ : Optional[int]="[CLS]" , lowerCamelCase__ : Tuple="[MASK]" , lowerCamelCase__ : int=True , lowerCamelCase__ : Optional[int]=None , **lowerCamelCase__ : str , ):
super().__init__(
lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , do_lower_case=lowerCamelCase__ , unk_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , tokenize_chinese_chars=lowerCamelCase__ , strip_accents=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : Optional[Any] = json.loads(self.backend_tokenizer.normalizer.__getstate__() )
if (
normalizer_state.get("lowercase" , lowerCamelCase__ ) != do_lower_case
or normalizer_state.get("strip_accents" , lowerCamelCase__ ) != strip_accents
or normalizer_state.get("handle_chinese_chars" , lowerCamelCase__ ) != tokenize_chinese_chars
):
a__ : Optional[Any] = getattr(lowerCamelCase__ , normalizer_state.pop("type" ) )
a__ : List[str] = do_lower_case
a__ : List[Any] = strip_accents
a__ : Optional[Any] = tokenize_chinese_chars
a__ : Optional[int] = normalizer_class(**lowerCamelCase__ )
a__ : Union[str, Any] = do_lower_case
def _UpperCamelCase( self : Any , lowerCamelCase__ : int , **lowerCamelCase__ : int ):
a__ : List[str] = PaddingStrategy.MAX_LENGTH
a__ : Dict = text
a__ : Optional[int] = kwargs.pop("text_pair" , lowerCamelCase__ )
a__ : Dict = kwargs.pop("return_tensors" , lowerCamelCase__ )
a__ : List[Any] = {
"input_ids": [],
"attention_mask": [],
"token_type_ids": [],
}
for idx, candidate_text in enumerate(lowerCamelCase__ ):
if batch_text_pair is not None:
a__ : Dict = batch_text_pair[idx]
else:
a__ : Union[str, Any] = None
a__ : int = super().__call__(lowerCamelCase__ , lowerCamelCase__ , return_tensors=lowerCamelCase__ , **lowerCamelCase__ )
a__ : Union[str, Any] = encoded_candidates.get("input_ids" )
a__ : List[Any] = encoded_candidates.get("attention_mask" )
a__ : int = encoded_candidates.get("token_type_ids" )
if encoded_input_ids is not None:
output_data["input_ids"].append(lowerCamelCase__ )
if encoded_attention_mask is not None:
output_data["attention_mask"].append(lowerCamelCase__ )
if encoded_token_type_ids is not None:
output_data["token_type_ids"].append(lowerCamelCase__ )
a__ : Union[str, Any] = {key: item for key, item in output_data.items() if len(lowerCamelCase__ ) != 0}
return BatchEncoding(lowerCamelCase__ , tensor_type=lowerCamelCase__ )
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : str=None ):
a__ : Optional[Any] = [self.cls_token_id] + token_ids_a + [self.sep_token_id]
if token_ids_a:
output += token_ids_a + [self.sep_token_id]
return output
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : List[Any] = [self.sep_token_id]
a__ : Optional[int] = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep ) * [0] + len(token_ids_a + sep ) * [1]
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : Tuple = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
| 37 |
import re
import jax.numpy as jnp
from flax.traverse_util import flatten_dict, unflatten_dict
from jax.random import PRNGKey
from ..utils import logging
UpperCamelCase : Dict = logging.get_logger(__name__)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
a__ : Tuple = R"\w+[.]\d+"
a__ : List[Any] = re.findall(__a , __a )
for pat in pats:
a__ : Union[str, Any] = key.replace(__a , "_".join(pat.split("." ) ) )
return key
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : List[str] = pt_tuple_key[:-1] + ("scale",)
if (
any("norm" in str_ for str_ in pt_tuple_key )
and (pt_tuple_key[-1] == "bias")
and (pt_tuple_key[:-1] + ("bias",) not in random_flax_state_dict)
and (pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict)
):
a__ : Any = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
elif pt_tuple_key[-1] in ["weight", "gamma"] and pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict:
a__ : Optional[Any] = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
# embedding
if pt_tuple_key[-1] == "weight" and pt_tuple_key[:-1] + ("embedding",) in random_flax_state_dict:
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("embedding",)
return renamed_pt_tuple_key, pt_tensor
# conv layer
a__ : List[str] = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight" and pt_tensor.ndim == 4:
a__ : str = pt_tensor.transpose(2 , 3 , 1 , 0 )
return renamed_pt_tuple_key, pt_tensor
# linear layer
a__ : Tuple = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight":
a__ : Tuple = pt_tensor.T
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm weight
a__ : Optional[Any] = pt_tuple_key[:-1] + ("weight",)
if pt_tuple_key[-1] == "gamma":
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm bias
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("bias",)
if pt_tuple_key[-1] == "beta":
return renamed_pt_tuple_key, pt_tensor
return pt_tuple_key, pt_tensor
def UpperCamelCase_ ( __a , __a , __a=42 ) -> str:
# Step 1: Convert pytorch tensor to numpy
a__ : Optional[int] = {k: v.numpy() for k, v in pt_state_dict.items()}
# Step 2: Since the model is stateless, get random Flax params
a__ : Tuple = flax_model.init_weights(PRNGKey(__a ) )
a__ : Optional[Any] = flatten_dict(__a )
a__ : Union[str, Any] = {}
# Need to change some parameters name to match Flax names
for pt_key, pt_tensor in pt_state_dict.items():
a__ : Optional[int] = rename_key(__a )
a__ : Optional[int] = tuple(renamed_pt_key.split("." ) )
# Correctly rename weight parameters
a__, a__ : Union[str, Any] = rename_key_and_reshape_tensor(__a , __a , __a )
if flax_key in random_flax_state_dict:
if flax_tensor.shape != random_flax_state_dict[flax_key].shape:
raise ValueError(
f'''PyTorch checkpoint seems to be incorrect. Weight {pt_key} was expected to be of shape '''
f'''{random_flax_state_dict[flax_key].shape}, but is {flax_tensor.shape}.''' )
# also add unexpected weight so that warning is thrown
a__ : str = jnp.asarray(__a )
return unflatten_dict(__a )
| 37 | 1 |
from dataclasses import dataclass
from typing import Optional, Tuple, Union
import numpy as np
import torch
from ..configuration_utils import ConfigMixin, register_to_config
from ..utils import BaseOutput, randn_tensor
from .scheduling_utils import SchedulerMixin
@dataclass
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 42
_lowercase = 42
_lowercase = None
class A__ ( A__ , A__ ):
"""simple docstring"""
_lowercase = 2
@register_to_config
def __init__( self : List[Any] , lowerCamelCase__ : float = 0.02 , lowerCamelCase__ : float = 100 , lowerCamelCase__ : float = 1.007 , lowerCamelCase__ : float = 80 , lowerCamelCase__ : float = 0.05 , lowerCamelCase__ : float = 50 , ):
# standard deviation of the initial noise distribution
a__ : Union[str, Any] = sigma_max
# setable values
a__ : int = None
a__ : np.IntTensor = None
a__ : torch.FloatTensor = None # sigma(t_i)
def _UpperCamelCase( self : Dict , lowerCamelCase__ : torch.FloatTensor , lowerCamelCase__ : Optional[int] = None ):
return sample
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : int , lowerCamelCase__ : Union[str, torch.device] = None ):
a__ : int = num_inference_steps
a__ : Optional[int] = np.arange(0 , self.num_inference_steps )[::-1].copy()
a__ : List[Any] = torch.from_numpy(lowerCamelCase__ ).to(lowerCamelCase__ )
a__ : Optional[int] = [
(
self.config.sigma_max**2
* (self.config.sigma_min**2 / self.config.sigma_max**2) ** (i / (num_inference_steps - 1))
)
for i in self.timesteps
]
a__ : Dict = torch.tensor(lowerCamelCase__ , dtype=torch.floataa , device=lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : torch.FloatTensor , lowerCamelCase__ : float , lowerCamelCase__ : Optional[torch.Generator] = None ):
if self.config.s_min <= sigma <= self.config.s_max:
a__ : Tuple = min(self.config.s_churn / self.num_inference_steps , 2**0.5 - 1 )
else:
a__ : Optional[int] = 0
# sample eps ~ N(0, S_noise^2 * I)
a__ : Optional[Any] = self.config.s_noise * randn_tensor(sample.shape , generator=lowerCamelCase__ ).to(sample.device )
a__ : str = sigma + gamma * sigma
a__ : List[str] = sample + ((sigma_hat**2 - sigma**2) ** 0.5 * eps)
return sample_hat, sigma_hat
def _UpperCamelCase( self : Dict , lowerCamelCase__ : torch.FloatTensor , lowerCamelCase__ : float , lowerCamelCase__ : float , lowerCamelCase__ : torch.FloatTensor , lowerCamelCase__ : bool = True , ):
a__ : str = sample_hat + sigma_hat * model_output
a__ : str = (sample_hat - pred_original_sample) / sigma_hat
a__ : Dict = sample_hat + (sigma_prev - sigma_hat) * derivative
if not return_dict:
return (sample_prev, derivative)
return KarrasVeOutput(
prev_sample=lowerCamelCase__ , derivative=lowerCamelCase__ , pred_original_sample=lowerCamelCase__ )
def _UpperCamelCase( self : Any , lowerCamelCase__ : torch.FloatTensor , lowerCamelCase__ : float , lowerCamelCase__ : float , lowerCamelCase__ : torch.FloatTensor , lowerCamelCase__ : torch.FloatTensor , lowerCamelCase__ : torch.FloatTensor , lowerCamelCase__ : bool = True , ):
a__ : int = sample_prev + sigma_prev * model_output
a__ : Any = (sample_prev - pred_original_sample) / sigma_prev
a__ : Tuple = sample_hat + (sigma_prev - sigma_hat) * (0.5 * derivative + 0.5 * derivative_corr)
if not return_dict:
return (sample_prev, derivative)
return KarrasVeOutput(
prev_sample=lowerCamelCase__ , derivative=lowerCamelCase__ , pred_original_sample=lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : int , lowerCamelCase__ : int ):
raise NotImplementedError()
| 37 |
from transformers import HfArgumentParser, TensorFlowBenchmark, TensorFlowBenchmarkArguments
def UpperCamelCase_ ( ) -> int:
a__ : Any = HfArgumentParser(__a )
a__ : Any = parser.parse_args_into_dataclasses()[0]
a__ : Optional[int] = TensorFlowBenchmark(args=__a )
try:
a__ : Optional[int] = parser.parse_args_into_dataclasses()[0]
except ValueError as e:
a__ : Tuple = "Arg --no_{0} is no longer used, please use --no-{0} instead."
a__ : List[Any] = " ".join(str(__a ).split(" " )[:-1] )
a__ : str = ""
a__ : List[Any] = eval(str(__a ).split(" " )[-1] )
a__ : List[str] = []
for arg in depreciated_args:
# arg[2:] removes '--'
if arg[2:] in TensorFlowBenchmark.deprecated_args:
# arg[5:] removes '--no_'
full_error_msg += arg_error_msg.format(arg[5:] )
else:
wrong_args.append(__a )
if len(__a ) > 0:
a__ : Tuple = full_error_msg + begin_error_msg + str(__a )
raise ValueError(__a )
benchmark.run()
if __name__ == "__main__":
main()
| 37 | 1 |
from typing import Any
class A__ :
"""simple docstring"""
def __init__( self : Union[str, Any] , lowerCamelCase__ : Any ):
a__ : List[str] = data
a__ : List[Any] = None
def __repr__( self : Tuple ):
return f'''Node({self.data})'''
class A__ :
"""simple docstring"""
def __init__( self : Dict ):
a__ : Union[str, Any] = None
def __iter__( self : List[Any] ):
a__ : List[str] = self.head
while node:
yield node.data
a__ : List[Any] = node.next
def __len__( self : Optional[Any] ):
return sum(1 for _ in self )
def __repr__( self : List[Any] ):
return "->".join([str(lowerCamelCase__ ) for item in self] )
def __getitem__( self : Dict , lowerCamelCase__ : int ):
if not 0 <= index < len(self ):
raise ValueError("list index out of range." )
for i, node in enumerate(self ):
if i == index:
return node
return None
def __setitem__( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any ):
if not 0 <= index < len(self ):
raise ValueError("list index out of range." )
a__ : int = self.head
for _ in range(lowerCamelCase__ ):
a__ : Union[str, Any] = current.next
a__ : Optional[int] = data
def _UpperCamelCase( self : Any , lowerCamelCase__ : Any ):
self.insert_nth(len(self ) , lowerCamelCase__ )
def _UpperCamelCase( self : str , lowerCamelCase__ : Any ):
self.insert_nth(0 , lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any ):
if not 0 <= index <= len(self ):
raise IndexError("list index out of range" )
a__ : Union[str, Any] = Node(lowerCamelCase__ )
if self.head is None:
a__ : List[Any] = new_node
elif index == 0:
a__ : str = self.head # link new_node to head
a__ : int = new_node
else:
a__ : List[Any] = self.head
for _ in range(index - 1 ):
a__ : Optional[int] = temp.next
a__ : Optional[int] = temp.next
a__ : Dict = new_node
def _UpperCamelCase( self : List[Any] ): # print every node data
print(self )
def _UpperCamelCase( self : List[str] ):
return self.delete_nth(0 )
def _UpperCamelCase( self : Dict ): # delete from tail
return self.delete_nth(len(self ) - 1 )
def _UpperCamelCase( self : Any , lowerCamelCase__ : int = 0 ):
if not 0 <= index <= len(self ) - 1: # test if index is valid
raise IndexError("List index out of range." )
a__ : Optional[int] = self.head # default first node
if index == 0:
a__ : Tuple = self.head.next
else:
a__ : List[Any] = self.head
for _ in range(index - 1 ):
a__ : int = temp.next
a__ : Tuple = temp.next
a__ : int = temp.next.next
return delete_node.data
def _UpperCamelCase( self : Optional[int] ):
return self.head is None
def _UpperCamelCase( self : Optional[int] ):
a__ : Any = None
a__ : List[Any] = self.head
while current:
# Store the current node's next node.
a__ : Tuple = current.next
# Make the current node's next point backwards
a__ : int = prev
# Make the previous node be the current node
a__ : Optional[Any] = current
# Make the current node the next node (to progress iteration)
a__ : Tuple = next_node
# Return prev in order to put the head at the end
a__ : Optional[Any] = prev
def UpperCamelCase_ ( ) -> None:
a__ : Tuple = LinkedList()
assert linked_list.is_empty() is True
assert str(__a ) == ""
try:
linked_list.delete_head()
raise AssertionError # This should not happen.
except IndexError:
assert True # This should happen.
try:
linked_list.delete_tail()
raise AssertionError # This should not happen.
except IndexError:
assert True # This should happen.
for i in range(10 ):
assert len(__a ) == i
linked_list.insert_nth(__a , i + 1 )
assert str(__a ) == "->".join(str(__a ) for i in range(1 , 11 ) )
linked_list.insert_head(0 )
linked_list.insert_tail(11 )
assert str(__a ) == "->".join(str(__a ) for i in range(0 , 12 ) )
assert linked_list.delete_head() == 0
assert linked_list.delete_nth(9 ) == 10
assert linked_list.delete_tail() == 11
assert len(__a ) == 9
assert str(__a ) == "->".join(str(__a ) for i in range(1 , 10 ) )
assert all(linked_list[i] == i + 1 for i in range(0 , 9 ) ) is True
for i in range(0 , 9 ):
a__ : int = -i
assert all(linked_list[i] == -i for i in range(0 , 9 ) ) is True
linked_list.reverse()
assert str(__a ) == "->".join(str(__a ) for i in range(-8 , 1 ) )
def UpperCamelCase_ ( ) -> None:
a__ : Union[str, Any] = [
-9,
100,
Node(77_345_112 ),
"dlrow olleH",
7,
5_555,
0,
-192.55555,
"Hello, world!",
77.9,
Node(10 ),
None,
None,
12.20,
]
a__ : Any = LinkedList()
for i in test_input:
linked_list.insert_tail(__a )
# Check if it's empty or not
assert linked_list.is_empty() is False
assert (
str(__a ) == "-9->100->Node(77345112)->dlrow olleH->7->5555->0->"
"-192.55555->Hello, world!->77.9->Node(10)->None->None->12.2"
)
# Delete the head
a__ : int = linked_list.delete_head()
assert result == -9
assert (
str(__a ) == "100->Node(77345112)->dlrow olleH->7->5555->0->-192.55555->"
"Hello, world!->77.9->Node(10)->None->None->12.2"
)
# Delete the tail
a__ : List[Any] = linked_list.delete_tail()
assert result == 12.2
assert (
str(__a ) == "100->Node(77345112)->dlrow olleH->7->5555->0->-192.55555->"
"Hello, world!->77.9->Node(10)->None->None"
)
# Delete a node in specific location in linked list
a__ : List[str] = linked_list.delete_nth(10 )
assert result is None
assert (
str(__a ) == "100->Node(77345112)->dlrow olleH->7->5555->0->-192.55555->"
"Hello, world!->77.9->Node(10)->None"
)
# Add a Node instance to its head
linked_list.insert_head(Node("Hello again, world!" ) )
assert (
str(__a )
== "Node(Hello again, world!)->100->Node(77345112)->dlrow olleH->"
"7->5555->0->-192.55555->Hello, world!->77.9->Node(10)->None"
)
# Add None to its tail
linked_list.insert_tail(__a )
assert (
str(__a )
== "Node(Hello again, world!)->100->Node(77345112)->dlrow olleH->"
"7->5555->0->-192.55555->Hello, world!->77.9->Node(10)->None->None"
)
# Reverse the linked list
linked_list.reverse()
assert (
str(__a )
== "None->None->Node(10)->77.9->Hello, world!->-192.55555->0->5555->"
"7->dlrow olleH->Node(77345112)->100->Node(Hello again, world!)"
)
def UpperCamelCase_ ( ) -> Tuple:
from doctest import testmod
testmod()
a__ : Any = LinkedList()
linked_list.insert_head(input("Inserting 1st at head " ).strip() )
linked_list.insert_head(input("Inserting 2nd at head " ).strip() )
print("\nPrint list:" )
linked_list.print_list()
linked_list.insert_tail(input("\nInserting 1st at tail " ).strip() )
linked_list.insert_tail(input("Inserting 2nd at tail " ).strip() )
print("\nPrint list:" )
linked_list.print_list()
print("\nDelete head" )
linked_list.delete_head()
print("Delete tail" )
linked_list.delete_tail()
print("\nPrint list:" )
linked_list.print_list()
print("\nReverse linked list" )
linked_list.reverse()
print("\nPrint list:" )
linked_list.print_list()
print("\nString representation of linked list:" )
print(__a )
print("\nReading/changing Node data using indexing:" )
print(f'''Element at Position 1: {linked_list[1]}''' )
a__ : Optional[int] = input("Enter New Value: " ).strip()
print("New list:" )
print(__a )
print(f'''length of linked_list is : {len(__a )}''' )
if __name__ == "__main__":
main()
| 37 |
import argparse
import ast
import logging
import os
import sys
import pandas as pd
import torch
from tqdm import tqdm
from transformers import BartForConditionalGeneration, RagRetriever, RagSequenceForGeneration, RagTokenForGeneration
from transformers import logging as transformers_logging
sys.path.append(os.path.join(os.getcwd())) # noqa: E402 # isort:skip
from utils_rag import exact_match_score, fa_score # noqa: E402 # isort:skip
UpperCamelCase : Optional[int] = logging.getLogger(__name__)
logging.basicConfig(level=logging.INFO)
transformers_logging.set_verbosity_info()
def UpperCamelCase_ ( __a ) -> Any:
if "token" in model_name_or_path:
return "rag_token"
if "sequence" in model_name_or_path:
return "rag_sequence"
if "bart" in model_name_or_path:
return "bart"
return None
def UpperCamelCase_ ( __a , __a , __a ) -> Any:
return max(metric_fn(__a , __a ) for gt in ground_truths )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = []
if args.gold_data_mode == "qa":
a__ : Any = pd.read_csv(__a , sep="\t" , header=__a )
for answer_list in data[1]:
a__ : Union[str, Any] = ast.literal_eval(__a )
answers.append(__a )
else:
a__ : List[str] = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : List[str] = [[reference] for reference in references]
a__ : List[str] = 0
for prediction, ground_truths in zip(__a , __a ):
total += 1
em += metric_max_over_ground_truths(__a , __a , __a )
fa += metric_max_over_ground_truths(__a , __a , __a )
a__ : Dict = 100.0 * em / total
a__ : Optional[Any] = 100.0 * fa / total
logger.info(f'''F1: {fa:.2f}''' )
logger.info(f'''EM: {em:.2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = args.k
a__ : str = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = 0
for hypo, reference in zip(__a , __a ):
a__ : Any = set(hypo.split("\t" )[:k] )
a__ : Union[str, Any] = set(reference.split("\t" ) )
total += 1
em += len(hypo_provenance & ref_provenance ) / k
a__ : Union[str, Any] = 100.0 * em / total
logger.info(f'''Precision@{k}: {em: .2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
def strip_title(__a ):
if title.startswith("\"" ):
a__ : Optional[Any] = title[1:]
if title.endswith("\"" ):
a__ : Union[str, Any] = title[:-1]
return title
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a , )["input_ids"].to(args.device )
a__ : Optional[int] = rag_model.rag.question_encoder(__a )
a__ : Union[str, Any] = question_enc_outputs[0]
a__ : Optional[int] = rag_model.retriever(
__a , question_enc_pool_output.cpu().detach().to(torch.floataa ).numpy() , prefix=rag_model.rag.generator.config.prefix , n_docs=rag_model.config.n_docs , return_tensors="pt" , )
a__ : List[Any] = rag_model.retriever.index.get_doc_dicts(result.doc_ids )
a__ : int = []
for docs in all_docs:
a__ : Optional[int] = [strip_title(__a ) for title in docs["title"]]
provenance_strings.append("\t".join(__a ) )
return provenance_strings
def UpperCamelCase_ ( __a , __a , __a ) -> Dict:
with torch.no_grad():
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a )
a__ : Any = inputs_dict.input_ids.to(args.device )
a__ : Dict = inputs_dict.attention_mask.to(args.device )
a__ : Optional[int] = rag_model.generate( # rag_model overwrites generate
__a , attention_mask=__a , num_beams=args.num_beams , min_length=args.min_length , max_length=args.max_length , early_stopping=__a , num_return_sequences=1 , bad_words_ids=[[0, 0]] , )
a__ : int = rag_model.retriever.generator_tokenizer.batch_decode(__a , skip_special_tokens=__a )
if args.print_predictions:
for q, a in zip(__a , __a ):
logger.info("Q: {} - A: {}".format(__a , __a ) )
return answers
def UpperCamelCase_ ( ) -> List[str]:
a__ : int = argparse.ArgumentParser()
parser.add_argument(
"--model_type" , choices=["rag_sequence", "rag_token", "bart"] , type=__a , help=(
"RAG model type: rag_sequence, rag_token or bart, if none specified, the type is inferred from the"
" model_name_or_path"
) , )
parser.add_argument(
"--index_name" , default=__a , choices=["exact", "compressed", "legacy"] , type=__a , help="RAG model retriever type" , )
parser.add_argument(
"--index_path" , default=__a , type=__a , help="Path to the retrieval index" , )
parser.add_argument("--n_docs" , default=5 , type=__a , help="Number of retrieved docs" )
parser.add_argument(
"--model_name_or_path" , default=__a , type=__a , required=__a , help="Path to pretrained checkpoints or model identifier from huggingface.co/models" , )
parser.add_argument(
"--eval_mode" , choices=["e2e", "retrieval"] , default="e2e" , type=__a , help=(
"Evaluation mode, e2e calculates exact match and F1 of the downstream task, retrieval calculates"
" precision@k."
) , )
parser.add_argument("--k" , default=1 , type=__a , help="k for the precision@k calculation" )
parser.add_argument(
"--evaluation_set" , default=__a , type=__a , required=__a , help="Path to a file containing evaluation samples" , )
parser.add_argument(
"--gold_data_path" , default=__a , type=__a , required=__a , help="Path to a tab-separated file with gold samples" , )
parser.add_argument(
"--gold_data_mode" , default="qa" , type=__a , choices=["qa", "ans"] , help=(
"Format of the gold data file"
"qa - a single line in the following format: question [tab] answer_list"
"ans - a single line of the gold file contains the expected answer string"
) , )
parser.add_argument(
"--predictions_path" , type=__a , default="predictions.txt" , help="Name of the predictions file, to be stored in the checkpoints directory" , )
parser.add_argument(
"--eval_all_checkpoints" , action="store_true" , help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number" , )
parser.add_argument(
"--eval_batch_size" , default=8 , type=__a , help="Batch size per GPU/CPU for evaluation." , )
parser.add_argument(
"--recalculate" , help="Recalculate predictions even if the prediction file exists" , action="store_true" , )
parser.add_argument(
"--num_beams" , default=4 , type=__a , help="Number of beams to be used when generating answers" , )
parser.add_argument("--min_length" , default=1 , type=__a , help="Min length of the generated answers" )
parser.add_argument("--max_length" , default=50 , type=__a , help="Max length of the generated answers" )
parser.add_argument(
"--print_predictions" , action="store_true" , help="If True, prints predictions while evaluating." , )
parser.add_argument(
"--print_docs" , action="store_true" , help="If True, prints docs retried while generating." , )
a__ : int = parser.parse_args()
a__ : Dict = torch.device("cuda" if torch.cuda.is_available() else "cpu" )
return args
def UpperCamelCase_ ( __a ) -> Optional[int]:
a__ : Tuple = {}
if args.model_type is None:
a__ : List[str] = infer_model_type(args.model_name_or_path )
assert args.model_type is not None
if args.model_type.startswith("rag" ):
a__ : int = RagTokenForGeneration if args.model_type == "rag_token" else RagSequenceForGeneration
a__ : Tuple = args.n_docs
if args.index_name is not None:
a__ : Any = args.index_name
if args.index_path is not None:
a__ : int = args.index_path
else:
a__ : Optional[Any] = BartForConditionalGeneration
a__ : Tuple = (
[f.path for f in os.scandir(args.model_name_or_path ) if f.is_dir()]
if args.eval_all_checkpoints
else [args.model_name_or_path]
)
logger.info("Evaluate the following checkpoints: %s" , __a )
a__ : Any = get_scores if args.eval_mode == "e2e" else get_precision_at_k
a__ : Union[str, Any] = evaluate_batch_eae if args.eval_mode == "e2e" else evaluate_batch_retrieval
for checkpoint in checkpoints:
if os.path.exists(args.predictions_path ) and (not args.recalculate):
logger.info("Calculating metrics based on an existing predictions file: {}".format(args.predictions_path ) )
score_fn(__a , args.predictions_path , args.gold_data_path )
continue
logger.info("***** Running evaluation for {} *****".format(__a ) )
logger.info(" Batch size = %d" , args.eval_batch_size )
logger.info(" Predictions will be stored under {}".format(args.predictions_path ) )
if args.model_type.startswith("rag" ):
a__ : str = RagRetriever.from_pretrained(__a , **__a )
a__ : Optional[int] = model_class.from_pretrained(__a , retriever=__a , **__a )
model.retriever.init_retrieval()
else:
a__ : Dict = model_class.from_pretrained(__a , **__a )
model.to(args.device )
with open(args.evaluation_set , "r" ) as eval_file, open(args.predictions_path , "w" ) as preds_file:
a__ : List[Any] = []
for line in tqdm(__a ):
questions.append(line.strip() )
if len(__a ) == args.eval_batch_size:
a__ : Union[str, Any] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) + "\n" )
preds_file.flush()
a__ : Any = []
if len(__a ) > 0:
a__ : List[str] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) )
preds_file.flush()
score_fn(__a , args.predictions_path , args.gold_data_path )
if __name__ == "__main__":
UpperCamelCase : List[Any] = get_args()
main(args)
| 37 | 1 |
def UpperCamelCase_ ( ) -> List[Any]:
a__ : Optional[int] = []
a__ : Dict = 1
while len(__a ) < 1e6:
constant.append(str(__a ) )
i += 1
a__ : Dict = "".join(__a )
return (
int(constant[0] )
* int(constant[9] )
* int(constant[99] )
* int(constant[999] )
* int(constant[9_999] )
* int(constant[99_999] )
* int(constant[999_999] )
)
if __name__ == "__main__":
print(solution())
| 37 |
import logging
import os
from dataclasses import dataclass, field
from typing import Dict, Optional
import datasets
import numpy as np
import tensorflow as tf
from transformers import (
AutoConfig,
AutoTokenizer,
EvalPrediction,
HfArgumentParser,
PreTrainedTokenizer,
TFAutoModelForSequenceClassification,
TFTrainer,
TFTrainingArguments,
)
from transformers.utils import logging as hf_logging
hf_logging.set_verbosity_info()
hf_logging.enable_default_handler()
hf_logging.enable_explicit_format()
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a = None , ) -> str:
a__ : int = {}
if train_file is not None:
a__ : int = [train_file]
if eval_file is not None:
a__ : Union[str, Any] = [eval_file]
if test_file is not None:
a__ : str = [test_file]
a__ : Optional[Any] = datasets.load_dataset("csv" , data_files=__a )
a__ : List[Any] = list(ds[list(files.keys() )[0]].features.keys() )
a__ : str = features_name.pop(__a )
a__ : Dict = list(set(ds[list(files.keys() )[0]][label_name] ) )
a__ : str = {label: i for i, label in enumerate(__a )}
a__ : Tuple = tokenizer.model_input_names
a__ : List[str] = {}
if len(__a ) == 1:
for k in files.keys():
a__ : Optional[Any] = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
example[features_name[0]] , truncation=__a , max_length=__a , padding="max_length" ) , batched=__a , )
elif len(__a ) == 2:
for k in files.keys():
a__ : Dict = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
(example[features_name[0]], example[features_name[1]]) , truncation=__a , max_length=__a , padding="max_length" , ) , batched=__a , )
def gen_train():
for ex in transformed_ds[datasets.Split.TRAIN]:
a__ : str = {k: v for k, v in ex.items() if k in input_names}
a__ : str = labelaid[ex[label_name]]
yield (d, label)
def gen_val():
for ex in transformed_ds[datasets.Split.VALIDATION]:
a__ : Tuple = {k: v for k, v in ex.items() if k in input_names}
a__ : List[Any] = labelaid[ex[label_name]]
yield (d, label)
def gen_test():
for ex in transformed_ds[datasets.Split.TEST]:
a__ : List[Any] = {k: v for k, v in ex.items() if k in input_names}
a__ : Optional[int] = labelaid[ex[label_name]]
yield (d, label)
a__ : Optional[Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TRAIN in transformed_ds
else None
)
if train_ds is not None:
a__ : Optional[int] = train_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TRAIN] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.VALIDATION in transformed_ds
else None
)
if val_ds is not None:
a__ : Optional[Any] = val_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.VALIDATION] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TEST in transformed_ds
else None
)
if test_ds is not None:
a__ : Tuple = test_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TEST] ) ) )
return train_ds, val_ds, test_ds, labelaid
UpperCamelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(metadata={'help': 'Which column contains the label'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the training file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the development file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the test file'} )
_lowercase = field(
default=1_2_8 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
_lowercase = field(
default=A__ , metadata={'help': 'Overwrite the cached training and evaluation sets'} )
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
_lowercase = field(default=A__ , metadata={'help': 'Set this flag to use fast tokenization.'} )
# If you want to tweak more attributes on your tokenizer, you should do it in a distinct script,
# or just modify its tokenizer_config.json.
_lowercase = field(
default=A__ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
def UpperCamelCase_ ( ) -> Union[str, Any]:
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
a__ : str = HfArgumentParser((ModelArguments, DataTrainingArguments, TFTrainingArguments) )
a__, a__, a__ : str = parser.parse_args_into_dataclasses()
if (
os.path.exists(training_args.output_dir )
and os.listdir(training_args.output_dir )
and training_args.do_train
and not training_args.overwrite_output_dir
):
raise ValueError(
f'''Output directory ({training_args.output_dir}) already exists and is not empty. Use'''
" --overwrite_output_dir to overcome." )
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s" , datefmt="%m/%d/%Y %H:%M:%S" , level=logging.INFO , )
logger.info(
f'''n_replicas: {training_args.n_replicas}, distributed training: {bool(training_args.n_replicas > 1 )}, '''
f'''16-bits training: {training_args.fpaa}''' )
logger.info(f'''Training/evaluation parameters {training_args}''' )
# Load pretrained model and tokenizer
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
a__ : Union[str, Any] = AutoTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , )
a__, a__, a__, a__ : Optional[Any] = get_tfds(
train_file=data_args.train_file , eval_file=data_args.dev_file , test_file=data_args.test_file , tokenizer=__a , label_column_id=data_args.label_column_id , max_seq_length=data_args.max_seq_length , )
a__ : Optional[int] = AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=len(__a ) , labelaid=__a , idalabel={id: label for label, id in labelaid.items()} , finetuning_task="text-classification" , cache_dir=model_args.cache_dir , )
with training_args.strategy.scope():
a__ : Any = TFAutoModelForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_pt=bool(".bin" in model_args.model_name_or_path ) , config=__a , cache_dir=model_args.cache_dir , )
def compute_metrics(__a ) -> Dict:
a__ : Union[str, Any] = np.argmax(p.predictions , axis=1 )
return {"acc": (preds == p.label_ids).mean()}
# Initialize our Trainer
a__ : Dict = TFTrainer(
model=__a , args=__a , train_dataset=__a , eval_dataset=__a , compute_metrics=__a , )
# Training
if training_args.do_train:
trainer.train()
trainer.save_model()
tokenizer.save_pretrained(training_args.output_dir )
# Evaluation
a__ : Optional[Any] = {}
if training_args.do_eval:
logger.info("*** Evaluate ***" )
a__ : Dict = trainer.evaluate()
a__ : int = os.path.join(training_args.output_dir , "eval_results.txt" )
with open(__a , "w" ) as writer:
logger.info("***** Eval results *****" )
for key, value in result.items():
logger.info(f''' {key} = {value}''' )
writer.write(f'''{key} = {value}\n''' )
results.update(__a )
return results
if __name__ == "__main__":
main()
| 37 | 1 |
from sklearn.metrics import mean_squared_error
import datasets
UpperCamelCase : Any = """\
@article{scikit-learn,
title={Scikit-learn: Machine Learning in {P}ython},
author={Pedregosa, F. and Varoquaux, G. and Gramfort, A. and Michel, V.
and Thirion, B. and Grisel, O. and Blondel, M. and Prettenhofer, P.
and Weiss, R. and Dubourg, V. and Vanderplas, J. and Passos, A. and
Cournapeau, D. and Brucher, M. and Perrot, M. and Duchesnay, E.},
journal={Journal of Machine Learning Research},
volume={12},
pages={2825--2830},
year={2011}
}
"""
UpperCamelCase : Optional[Any] = """\
Mean Squared Error(MSE) is the average of the square of difference between the predicted
and actual values.
"""
UpperCamelCase : Optional[Any] = """
Args:
predictions: array-like of shape (n_samples,) or (n_samples, n_outputs)
Estimated target values.
references: array-like of shape (n_samples,) or (n_samples, n_outputs)
Ground truth (correct) target values.
sample_weight: array-like of shape (n_samples,), default=None
Sample weights.
multioutput: {\"raw_values\", \"uniform_average\"} or array-like of shape (n_outputs,), default=\"uniform_average\"
Defines aggregating of multiple output values. Array-like value defines weights used to average errors.
\"raw_values\" : Returns a full set of errors in case of multioutput input.
\"uniform_average\" : Errors of all outputs are averaged with uniform weight.
squared : bool, default=True
If True returns MSE value, if False returns RMSE (Root Mean Squared Error) value.
Returns:
mse : mean squared error.
Examples:
>>> mse_metric = datasets.load_metric(\"mse\")
>>> predictions = [2.5, 0.0, 2, 8]
>>> references = [3, -0.5, 2, 7]
>>> results = mse_metric.compute(predictions=predictions, references=references)
>>> print(results)
{'mse': 0.375}
>>> rmse_result = mse_metric.compute(predictions=predictions, references=references, squared=False)
>>> print(rmse_result)
{'mse': 0.6123724356957945}
If you're using multi-dimensional lists, then set the config as follows :
>>> mse_metric = datasets.load_metric(\"mse\", \"multilist\")
>>> predictions = [[0.5, 1], [-1, 1], [7, -6]]
>>> references = [[0, 2], [-1, 2], [8, -5]]
>>> results = mse_metric.compute(predictions=predictions, references=references)
>>> print(results)
{'mse': 0.7083333333333334}
>>> results = mse_metric.compute(predictions=predictions, references=references, multioutput='raw_values')
>>> print(results) # doctest: +NORMALIZE_WHITESPACE
{'mse': array([0.41666667, 1. ])}
"""
@datasets.utils.file_utils.add_start_docstrings(_DESCRIPTION , _KWARGS_DESCRIPTION )
class A__ ( datasets.Metric ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
return datasets.MetricInfo(
description=_DESCRIPTION , citation=_CITATION , inputs_description=_KWARGS_DESCRIPTION , features=datasets.Features(self._get_feature_types() ) , reference_urls=[
"https://scikit-learn.org/stable/modules/generated/sklearn.metrics.mean_squared_error.html"
] , )
def _UpperCamelCase( self : Optional[int] ):
if self.config_name == "multilist":
return {
"predictions": datasets.Sequence(datasets.Value("float" ) ),
"references": datasets.Sequence(datasets.Value("float" ) ),
}
else:
return {
"predictions": datasets.Value("float" ),
"references": datasets.Value("float" ),
}
def _UpperCamelCase( self : int , lowerCamelCase__ : List[str] , lowerCamelCase__ : List[str] , lowerCamelCase__ : Optional[Any]=None , lowerCamelCase__ : Dict="uniform_average" , lowerCamelCase__ : Tuple=True ):
a__ : int = mean_squared_error(
lowerCamelCase__ , lowerCamelCase__ , sample_weight=lowerCamelCase__ , multioutput=lowerCamelCase__ , squared=lowerCamelCase__ )
return {"mse": mse}
| 37 |
import argparse
import collections
import json
import os
import re
import string
import sys
import numpy as np
UpperCamelCase : List[str] = re.compile(r"""\b(a|an|the)\b""", re.UNICODE)
UpperCamelCase : Union[str, Any] = None
def UpperCamelCase_ ( ) -> List[str]:
a__ : List[Any] = argparse.ArgumentParser("Official evaluation script for SQuAD version 2.0." )
parser.add_argument("data_file" , metavar="data.json" , help="Input data JSON file." )
parser.add_argument("pred_file" , metavar="pred.json" , help="Model predictions." )
parser.add_argument(
"--out-file" , "-o" , metavar="eval.json" , help="Write accuracy metrics to file (default is stdout)." )
parser.add_argument(
"--na-prob-file" , "-n" , metavar="na_prob.json" , help="Model estimates of probability of no answer." )
parser.add_argument(
"--na-prob-thresh" , "-t" , type=__a , default=1.0 , help="Predict \"\" if no-answer probability exceeds this (default = 1.0)." , )
parser.add_argument(
"--out-image-dir" , "-p" , metavar="out_images" , default=__a , help="Save precision-recall curves to directory." )
parser.add_argument("--verbose" , "-v" , action="store_true" )
if len(sys.argv ) == 1:
parser.print_help()
sys.exit(1 )
return parser.parse_args()
def UpperCamelCase_ ( __a ) -> str:
a__ : Optional[Any] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : Dict = bool(qa["answers"]["text"] )
return qid_to_has_ans
def UpperCamelCase_ ( __a ) -> List[Any]:
def remove_articles(__a ):
return ARTICLES_REGEX.sub(" " , __a )
def white_space_fix(__a ):
return " ".join(text.split() )
def remove_punc(__a ):
a__ : Union[str, Any] = set(string.punctuation )
return "".join(ch for ch in text if ch not in exclude )
def lower(__a ):
return text.lower()
return white_space_fix(remove_articles(remove_punc(lower(__a ) ) ) )
def UpperCamelCase_ ( __a ) -> Dict:
if not s:
return []
return normalize_answer(__a ).split()
def UpperCamelCase_ ( __a , __a ) -> str:
return int(normalize_answer(__a ) == normalize_answer(__a ) )
def UpperCamelCase_ ( __a , __a ) -> Dict:
a__ : int = get_tokens(__a )
a__ : Optional[Any] = get_tokens(__a )
a__ : Any = collections.Counter(__a ) & collections.Counter(__a )
a__ : Dict = sum(common.values() )
if len(__a ) == 0 or len(__a ) == 0:
# If either is no-answer, then F1 is 1 if they agree, 0 otherwise
return int(gold_toks == pred_toks )
if num_same == 0:
return 0
a__ : Tuple = 1.0 * num_same / len(__a )
a__ : str = 1.0 * num_same / len(__a )
a__ : str = (2 * precision * recall) / (precision + recall)
return fa
def UpperCamelCase_ ( __a , __a ) -> int:
a__ : List[str] = {}
a__ : Optional[int] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : List[Any] = qa["id"]
a__ : Dict = [t for t in qa["answers"]["text"] if normalize_answer(__a )]
if not gold_answers:
# For unanswerable questions, only correct answer is empty string
a__ : Tuple = [""]
if qid not in preds:
print(f'''Missing prediction for {qid}''' )
continue
a__ : Tuple = preds[qid]
# Take max over all gold answers
a__ : Optional[int] = max(compute_exact(__a , __a ) for a in gold_answers )
a__ : str = max(compute_fa(__a , __a ) for a in gold_answers )
return exact_scores, fa_scores
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
a__ : Optional[Any] = {}
for qid, s in scores.items():
a__ : Dict = na_probs[qid] > na_prob_thresh
if pred_na:
a__ : Dict = float(not qid_to_has_ans[qid] )
else:
a__ : Optional[Any] = s
return new_scores
def UpperCamelCase_ ( __a , __a , __a=None ) -> Tuple:
if not qid_list:
a__ : Union[str, Any] = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores.values() ) / total),
("f1", 100.0 * sum(fa_scores.values() ) / total),
("total", total),
] )
else:
a__ : int = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores[k] for k in qid_list ) / total),
("f1", 100.0 * sum(fa_scores[k] for k in qid_list ) / total),
("total", total),
] )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
for k in new_eval:
a__ : Optional[Any] = new_eval[k]
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
plt.step(__a , __a , color="b" , alpha=0.2 , where="post" )
plt.fill_between(__a , __a , step="post" , alpha=0.2 , color="b" )
plt.xlabel("Recall" )
plt.ylabel("Precision" )
plt.xlim([0.0, 1.05] )
plt.ylim([0.0, 1.05] )
plt.title(__a )
plt.savefig(__a )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a , __a=None , __a=None ) -> Dict:
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
a__ : Any = 0.0
a__ : Optional[int] = 1.0
a__ : Optional[int] = 0.0
a__ : Any = [1.0]
a__ : Tuple = [0.0]
a__ : List[str] = 0.0
for i, qid in enumerate(__a ):
if qid_to_has_ans[qid]:
true_pos += scores[qid]
a__ : Any = true_pos / float(i + 1 )
a__ : int = true_pos / float(__a )
if i == len(__a ) - 1 or na_probs[qid] != na_probs[qid_list[i + 1]]:
# i.e., if we can put a threshold after this point
avg_prec += cur_p * (cur_r - recalls[-1])
precisions.append(__a )
recalls.append(__a )
if out_image:
plot_pr_curve(__a , __a , __a , __a )
return {"ap": 100.0 * avg_prec}
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> str:
if out_image_dir and not os.path.exists(__a ):
os.makedirs(__a )
a__ : Optional[int] = sum(1 for v in qid_to_has_ans.values() if v )
if num_true_pos == 0:
return
a__ : Optional[int] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_exact.png" ) , title="Precision-Recall curve for Exact Match score" , )
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_f1.png" ) , title="Precision-Recall curve for F1 score" , )
a__ : str = {k: float(__a ) for k, v in qid_to_has_ans.items()}
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_oracle.png" ) , title="Oracle Precision-Recall curve (binary task of HasAns vs. NoAns)" , )
merge_eval(__a , __a , "pr_exact" )
merge_eval(__a , __a , "pr_f1" )
merge_eval(__a , __a , "pr_oracle" )
def UpperCamelCase_ ( __a , __a , __a , __a ) -> str:
if not qid_list:
return
a__ : Optional[Any] = [na_probs[k] for k in qid_list]
a__ : str = np.ones_like(__a ) / float(len(__a ) )
plt.hist(__a , weights=__a , bins=20 , range=(0.0, 1.0) )
plt.xlabel("Model probability of no-answer" )
plt.ylabel("Proportion of dataset" )
plt.title(f'''Histogram of no-answer probability: {name}''' )
plt.savefig(os.path.join(__a , f'''na_prob_hist_{name}.png''' ) )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[Any]:
a__ : str = sum(1 for k in qid_to_has_ans if not qid_to_has_ans[k] )
a__ : Optional[Any] = num_no_ans
a__ : Dict = cur_score
a__ : Any = 0.0
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
for i, qid in enumerate(__a ):
if qid not in scores:
continue
if qid_to_has_ans[qid]:
a__ : Optional[int] = scores[qid]
else:
if preds[qid]:
a__ : str = -1
else:
a__ : Union[str, Any] = 0
cur_score += diff
if cur_score > best_score:
a__ : Any = cur_score
a__ : Dict = na_probs[qid]
return 100.0 * best_score / len(__a ), best_thresh
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> Any:
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__ : Any = best_exact
a__ : Any = exact_thresh
a__ : List[Any] = best_fa
a__ : Optional[int] = fa_thresh
def UpperCamelCase_ ( ) -> Tuple:
with open(OPTS.data_file ) as f:
a__ : List[Any] = json.load(__a )
a__ : Any = dataset_json["data"]
with open(OPTS.pred_file ) as f:
a__ : int = json.load(__a )
if OPTS.na_prob_file:
with open(OPTS.na_prob_file ) as f:
a__ : List[str] = json.load(__a )
else:
a__ : Optional[int] = {k: 0.0 for k in preds}
a__ : Optional[Any] = make_qid_to_has_ans(__a ) # maps qid to True/False
a__ : List[Any] = [k for k, v in qid_to_has_ans.items() if v]
a__ : Union[str, Any] = [k for k, v in qid_to_has_ans.items() if not v]
a__, a__ : str = get_raw_scores(__a , __a )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : Tuple = make_eval_dict(__a , __a )
if has_ans_qids:
a__ : str = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "HasAns" )
if no_ans_qids:
a__ : List[Any] = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "NoAns" )
if OPTS.na_prob_file:
find_all_best_thresh(__a , __a , __a , __a , __a , __a )
if OPTS.na_prob_file and OPTS.out_image_dir:
run_precision_recall_analysis(__a , __a , __a , __a , __a , OPTS.out_image_dir )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "hasAns" )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "noAns" )
if OPTS.out_file:
with open(OPTS.out_file , "w" ) as f:
json.dump(__a , __a )
else:
print(json.dumps(__a , indent=2 ) )
if __name__ == "__main__":
UpperCamelCase : Any = parse_args()
if OPTS.out_image_dir:
import matplotlib
matplotlib.use("""Agg""")
import matplotlib.pyplot as plt
main()
| 37 | 1 |
from datasets.utils.patching import _PatchedModuleObj, patch_submodule
from . import _test_patching
def UpperCamelCase_ ( ) -> Optional[int]:
import os as original_os
from os import path as original_path
from os import rename as original_rename
from os.path import dirname as original_dirname
from os.path import join as original_join
assert _test_patching.os is original_os
assert _test_patching.path is original_path
assert _test_patching.join is original_join
assert _test_patching.renamed_os is original_os
assert _test_patching.renamed_path is original_path
assert _test_patching.renamed_join is original_join
a__ : Dict = "__test_patch_submodule_mock__"
with patch_submodule(_test_patching , "os.path.join" , __a ):
# Every way to access os.path.join must be patched, and the rest must stay untouched
# check os.path.join
assert isinstance(_test_patching.os , _PatchedModuleObj )
assert isinstance(_test_patching.os.path , _PatchedModuleObj )
assert _test_patching.os.path.join is mock
# check path.join
assert isinstance(_test_patching.path , _PatchedModuleObj )
assert _test_patching.path.join is mock
# check join
assert _test_patching.join is mock
# check that the other attributes are untouched
assert _test_patching.os.rename is original_rename
assert _test_patching.path.dirname is original_dirname
assert _test_patching.os.path.dirname is original_dirname
# Even renamed modules or objects must be patched
# check renamed_os.path.join
assert isinstance(_test_patching.renamed_os , _PatchedModuleObj )
assert isinstance(_test_patching.renamed_os.path , _PatchedModuleObj )
assert _test_patching.renamed_os.path.join is mock
# check renamed_path.join
assert isinstance(_test_patching.renamed_path , _PatchedModuleObj )
assert _test_patching.renamed_path.join is mock
# check renamed_join
assert _test_patching.renamed_join is mock
# check that the other attributes are untouched
assert _test_patching.renamed_os.rename is original_rename
assert _test_patching.renamed_path.dirname is original_dirname
assert _test_patching.renamed_os.path.dirname is original_dirname
# check that everthing is back to normal when the patch is over
assert _test_patching.os is original_os
assert _test_patching.path is original_path
assert _test_patching.join is original_join
assert _test_patching.renamed_os is original_os
assert _test_patching.renamed_path is original_path
assert _test_patching.renamed_join is original_join
def UpperCamelCase_ ( ) -> Optional[Any]:
assert _test_patching.open is open
a__ : Dict = "__test_patch_submodule_builtin_mock__"
# _test_patching has "open" in its globals
assert _test_patching.open is open
with patch_submodule(_test_patching , "open" , __a ):
assert _test_patching.open is mock
# check that everthing is back to normal when the patch is over
assert _test_patching.open is open
def UpperCamelCase_ ( ) -> Tuple:
# pandas.read_csv is not present in _test_patching
a__ : List[str] = "__test_patch_submodule_missing_mock__"
with patch_submodule(_test_patching , "pandas.read_csv" , __a ):
pass
def UpperCamelCase_ ( ) -> int:
# builtin should always be mocked even if they're not in the globals
# in case they're loaded at one point
a__ : List[str] = "__test_patch_submodule_missing_builtin_mock__"
# _test_patching doesn't have "len" in its globals
assert getattr(_test_patching , "len" , __a ) is None
with patch_submodule(_test_patching , "len" , __a ):
assert _test_patching.len is mock
assert _test_patching.len is len
def UpperCamelCase_ ( ) -> List[str]:
a__ : Tuple = "__test_patch_submodule_start_and_stop_mock__"
a__ : Dict = patch_submodule(_test_patching , "open" , __a )
assert _test_patching.open is open
patch.start()
assert _test_patching.open is mock
patch.stop()
assert _test_patching.open is open
def UpperCamelCase_ ( ) -> Union[str, Any]:
from os import rename as original_rename
from os.path import dirname as original_dirname
from os.path import join as original_join
a__ : List[Any] = "__test_patch_submodule_successive_join__"
a__ : Tuple = "__test_patch_submodule_successive_dirname__"
a__ : Dict = "__test_patch_submodule_successive_rename__"
assert _test_patching.os.path.join is original_join
assert _test_patching.os.path.dirname is original_dirname
assert _test_patching.os.rename is original_rename
with patch_submodule(_test_patching , "os.path.join" , __a ):
with patch_submodule(_test_patching , "os.rename" , __a ):
with patch_submodule(_test_patching , "os.path.dirname" , __a ):
assert _test_patching.os.path.join is mock_join
assert _test_patching.os.path.dirname is mock_dirname
assert _test_patching.os.rename is mock_rename
# try another order
with patch_submodule(_test_patching , "os.rename" , __a ):
with patch_submodule(_test_patching , "os.path.join" , __a ):
with patch_submodule(_test_patching , "os.path.dirname" , __a ):
assert _test_patching.os.path.join is mock_join
assert _test_patching.os.path.dirname is mock_dirname
assert _test_patching.os.rename is mock_rename
assert _test_patching.os.path.join is original_join
assert _test_patching.os.path.dirname is original_dirname
assert _test_patching.os.rename is original_rename
def UpperCamelCase_ ( ) -> List[str]:
a__ : Tuple = "__test_patch_submodule_doesnt_exist_mock__"
with patch_submodule(_test_patching , "__module_that_doesn_exist__.__attribute_that_doesn_exist__" , __a ):
pass
with patch_submodule(_test_patching , "os.__attribute_that_doesn_exist__" , __a ):
pass
| 37 |
import json
import os
import unittest
from transformers import CLIPTokenizer, CLIPTokenizerFast
from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_ftfy, require_tokenizers
from ...test_tokenization_common import TokenizerTesterMixin
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = CLIPTokenizer
_lowercase = CLIPTokenizerFast
_lowercase = True
_lowercase = {}
_lowercase = False
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# fmt: off
a__ : Any = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "lo", "l</w>", "w</w>", "r</w>", "t</w>", "low</w>", "er</w>", "lowest</w>", "newer</w>", "wider", "<unk>", "<|startoftext|>", "<|endoftext|>"]
# fmt: on
a__ : Optional[Any] = dict(zip(lowerCamelCase__ , range(len(lowerCamelCase__ ) ) ) )
a__ : Optional[Any] = ["#version: 0.2", "l o", "lo w</w>", "e r</w>"]
a__ : Optional[Any] = {"unk_token": "<unk>"}
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["vocab_file"] )
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["merges_file"] )
with open(self.vocab_file , "w" , encoding="utf-8" ) as fp:
fp.write(json.dumps(lowerCamelCase__ ) + "\n" )
with open(self.merges_file , "w" , encoding="utf-8" ) as fp:
fp.write("\n".join(lowerCamelCase__ ) )
def _UpperCamelCase( self : Dict , **lowerCamelCase__ : int ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizer.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , **lowerCamelCase__ : Optional[int] ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizerFast.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[Any] ):
a__ : int = "lower newer"
a__ : Optional[int] = "lower newer"
return input_text, output_text
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = CLIPTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
a__ : int = "lower newer"
a__ : List[str] = ["lo", "w", "er</w>", "n", "e", "w", "er</w>"]
a__ : Union[str, Any] = tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : int = tokens + [tokenizer.unk_token]
a__ : Union[str, Any] = [10, 2, 16, 9, 3, 2, 16, 20]
self.assertListEqual(tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , lowerCamelCase__ )
@require_ftfy
def _UpperCamelCase( self : Optional[Any] ):
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : List[str] = self.tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : Any = self.rust_tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : int = "A\n'll 11p223RF☆ho!!to?'d'd''d of a cat to-$''d."
a__ : Optional[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : Dict = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on an example containing a character (Latin Small Letter A
# with Tilde) encoded in 2 different ways
a__ : Optional[Any] = "xa\u0303y" + " " + "x\xe3y"
a__ : Optional[int] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of space type
a__ : str = [
"\u0009", # (horizontal tab, '\t')
"\u000B", # (vertical tab)
"\u000C", # (form feed)
"\u0020", # (space, ' ')
"\u200E", # (left-to-right mark):w
"\u200F", # (right-to-left mark)
]
for unicode_seq in spaces_unicodes:
a__ : Any = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of line break type
a__ : Union[str, Any] = [
"\u000A", # (line feed, '\n')
"\r\n", # (carriage return and line feed, '\r\n')
"\u000D", # (carriage return, '\r')
"\r", # (carriage return, '\r')
"\u000D", # (carriage return, '\r')
"\u2028", # (line separator)
"\u2029", # (paragraph separator)
# "\u0085", # (next line)
]
# The tokenization is not identical for the character "\u0085" (next line). The slow version using ftfy transforms
# it into the Horizontal Ellipsis character "…" ("\u2026") while the fast version transforms it into a
# space (and thus into an empty list).
for unicode_seq in line_break_unicodes:
a__ : List[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
# Test which aims to verify that the offsets are well adapted to the argument `add_prefix_space`
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : str = "hello" # `hello` is a token in the vocabulary of `pretrained_name`
a__ : Tuple = f'''{text_of_1_token} {text_of_1_token}'''
a__ : Optional[int] = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Union[str, Any] = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (0, len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (len(lowerCamelCase__ ) + 1, len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
a__ : Optional[Any] = f''' {text}'''
a__ : str = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Dict = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (1, 1 + len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (1 + len(lowerCamelCase__ ) + 1, 1 + len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
def _UpperCamelCase( self : int ):
# Test related to the breaking change introduced in transformers v4.17.0
# We need to check that an error in raised when the user try to load a previous version of the tokenizer.
with self.assertRaises(lowerCamelCase__ ) as context:
self.rust_tokenizer_class.from_pretrained("robot-test/old-clip-tokenizer" )
self.assertTrue(
context.exception.args[0].startswith(
"The `backend_tokenizer` provided does not match the expected format." ) )
@require_ftfy
def _UpperCamelCase( self : int ):
super().test_tokenization_python_rust_equals()
def _UpperCamelCase( self : str ):
# CLIP always lower cases letters
pass
| 37 | 1 |
from ...configuration_utils import PretrainedConfig
from ...utils import logging
UpperCamelCase : Any = logging.get_logger(__name__)
UpperCamelCase : Union[str, Any] = {
"""facebook/xglm-564M""": """https://huggingface.co/facebook/xglm-564M/resolve/main/config.json""",
# See all XGLM models at https://huggingface.co/models?filter=xglm
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'xglm'
_lowercase = ['past_key_values']
_lowercase = {
'num_attention_heads': 'attention_heads',
'hidden_size': 'd_model',
'num_hidden_layers': 'num_layers',
}
def __init__( self : List[str] , lowerCamelCase__ : Optional[Any]=256_008 , lowerCamelCase__ : Dict=2_048 , lowerCamelCase__ : Optional[int]=1_024 , lowerCamelCase__ : Any=4_096 , lowerCamelCase__ : Any=24 , lowerCamelCase__ : Tuple=16 , lowerCamelCase__ : str="gelu" , lowerCamelCase__ : List[str]=0.1 , lowerCamelCase__ : Optional[int]=0.1 , lowerCamelCase__ : int=0.0 , lowerCamelCase__ : List[str]=0.0 , lowerCamelCase__ : Optional[Any]=0.02 , lowerCamelCase__ : Tuple=True , lowerCamelCase__ : Dict=True , lowerCamelCase__ : List[str]=2 , lowerCamelCase__ : Dict=1 , lowerCamelCase__ : Optional[int]=0 , lowerCamelCase__ : int=2 , **lowerCamelCase__ : Optional[int] , ):
a__ : int = vocab_size
a__ : int = max_position_embeddings
a__ : Union[str, Any] = d_model
a__ : List[Any] = ffn_dim
a__ : Optional[int] = num_layers
a__ : int = attention_heads
a__ : Any = activation_function
a__ : Union[str, Any] = dropout
a__ : List[Any] = attention_dropout
a__ : Union[str, Any] = activation_dropout
a__ : Union[str, Any] = layerdrop
a__ : Any = init_std
a__ : int = scale_embedding # scale factor will be sqrt(d_model) if True
a__ : List[str] = use_cache
super().__init__(
pad_token_id=lowerCamelCase__ , bos_token_id=lowerCamelCase__ , eos_token_id=lowerCamelCase__ , decoder_start_token_id=lowerCamelCase__ , **lowerCamelCase__ , )
| 37 |
import os
import re
import shutil
from argparse import ArgumentParser, Namespace
from datasets.commands import BaseDatasetsCLICommand
from datasets.utils.logging import get_logger
UpperCamelCase : Dict = """<<<<<<< This should probably be modified because it mentions: """
UpperCamelCase : List[Any] = """=======
>>>>>>>
"""
UpperCamelCase : Optional[Any] = [
"""TextEncoderConfig""",
"""ByteTextEncoder""",
"""SubwordTextEncoder""",
"""encoder_config""",
"""maybe_build_from_corpus""",
"""manual_dir""",
]
UpperCamelCase : Any = [
# (pattern, replacement)
# Order is important here for some replacements
(r"""tfds\.core""", r"""datasets"""),
(r"""tf\.io\.gfile\.GFile""", r"""open"""),
(r"""tf\.([\w\d]+)""", r"""datasets.Value('\1')"""),
(r"""tfds\.features\.Text\(\)""", r"""datasets.Value('string')"""),
(r"""tfds\.features\.Text\(""", r"""datasets.Value('string'),"""),
(r"""features\s*=\s*tfds.features.FeaturesDict\(""", r"""features=datasets.Features("""),
(r"""tfds\.features\.FeaturesDict\(""", r"""dict("""),
(r"""The TensorFlow Datasets Authors""", r"""The TensorFlow Datasets Authors and the HuggingFace Datasets Authors"""),
(r"""tfds\.""", r"""datasets."""),
(r"""dl_manager\.manual_dir""", r"""self.config.data_dir"""),
(r"""self\.builder_config""", r"""self.config"""),
]
def UpperCamelCase_ ( __a ) -> Optional[Any]:
return ConvertCommand(args.tfds_path , args.datasets_directory )
class A__ ( A__ ):
"""simple docstring"""
@staticmethod
def _UpperCamelCase( lowerCamelCase__ : ArgumentParser ):
a__ : List[str] = parser.add_parser(
"convert" , help="Convert a TensorFlow Datasets dataset to a HuggingFace Datasets dataset." , )
train_parser.add_argument(
"--tfds_path" , type=lowerCamelCase__ , required=lowerCamelCase__ , help="Path to a TensorFlow Datasets folder to convert or a single tfds file to convert." , )
train_parser.add_argument(
"--datasets_directory" , type=lowerCamelCase__ , required=lowerCamelCase__ , help="Path to the HuggingFace Datasets folder." )
train_parser.set_defaults(func=lowerCamelCase__ )
def __init__( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str , *lowerCamelCase__ : Tuple ):
a__ : str = get_logger("datasets-cli/converting" )
a__ : Optional[Any] = tfds_path
a__ : Optional[int] = datasets_directory
def _UpperCamelCase( self : int ):
if os.path.isdir(self._tfds_path ):
a__ : List[str] = os.path.abspath(self._tfds_path )
elif os.path.isfile(self._tfds_path ):
a__ : Any = os.path.dirname(self._tfds_path )
else:
raise ValueError("--tfds_path is neither a directory nor a file. Please check path." )
a__ : Dict = os.path.abspath(self._datasets_directory )
self._logger.info(f'''Converting datasets from {abs_tfds_path} to {abs_datasets_path}''' )
a__ : Tuple = []
a__ : str = []
a__ : List[Any] = {}
if os.path.isdir(self._tfds_path ):
a__ : List[str] = os.listdir(lowerCamelCase__ )
else:
a__ : Union[str, Any] = [os.path.basename(self._tfds_path )]
for f_name in file_names:
self._logger.info(f'''Looking at file {f_name}''' )
a__ : Any = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
a__ : Dict = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
if not os.path.isfile(lowerCamelCase__ ) or "__init__" in f_name or "_test" in f_name or ".py" not in f_name:
self._logger.info("Skipping file" )
continue
with open(lowerCamelCase__ , encoding="utf-8" ) as f:
a__ : List[Any] = f.readlines()
a__ : Union[str, Any] = []
a__ : Union[str, Any] = False
a__ : Union[str, Any] = False
a__ : Dict = []
for line in lines:
a__ : Optional[Any] = line
# Convert imports
if "import tensorflow.compat.v2 as tf" in out_line:
continue
elif "@tfds.core" in out_line:
continue
elif "builder=self" in out_line:
continue
elif "import tensorflow_datasets.public_api as tfds" in out_line:
a__ : List[Any] = "import datasets\n"
elif "import tensorflow" in out_line:
# order is important here
a__ : List[str] = ""
continue
elif "from absl import logging" in out_line:
a__ : Dict = "from datasets import logging\n"
elif "getLogger" in out_line:
a__ : List[Any] = out_line.replace("getLogger" , "get_logger" )
elif any(expression in out_line for expression in TO_HIGHLIGHT ):
a__ : List[str] = True
a__ : Dict = list(filter(lambda lowerCamelCase__ : e in out_line , lowerCamelCase__ ) )
out_lines.append(HIGHLIGHT_MESSAGE_PRE + str(lowerCamelCase__ ) + "\n" )
out_lines.append(lowerCamelCase__ )
out_lines.append(lowerCamelCase__ )
continue
else:
for pattern, replacement in TO_CONVERT:
a__ : Tuple = re.sub(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
# Take care of saving utilities (to later move them together with main script)
if "tensorflow_datasets" in out_line:
a__ : Optional[int] = re.match(r"from\stensorflow_datasets.*import\s([^\.\r\n]+)" , lowerCamelCase__ )
tfds_imports.extend(imp.strip() for imp in match.group(1 ).split("," ) )
a__ : Optional[Any] = "from . import " + match.group(1 )
# Check we have not forget anything
if "tf." in out_line or "tfds." in out_line or "tensorflow_datasets" in out_line:
raise ValueError(f'''Error converting {out_line.strip()}''' )
if "GeneratorBasedBuilder" in out_line or "BeamBasedBuilder" in out_line:
a__ : Optional[int] = True
out_lines.append(lowerCamelCase__ )
if is_builder or "wmt" in f_name:
# We create a new directory for each dataset
a__ : Dict = f_name.replace(".py" , "" )
a__ : Optional[int] = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
a__ : Any = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
os.makedirs(lowerCamelCase__ , exist_ok=lowerCamelCase__ )
self._logger.info(f'''Adding directory {output_dir}''' )
imports_to_builder_map.update({imp: output_dir for imp in tfds_imports} )
else:
# Utilities will be moved at the end
utils_files.append(lowerCamelCase__ )
if needs_manual_update:
with_manual_update.append(lowerCamelCase__ )
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as f:
f.writelines(lowerCamelCase__ )
self._logger.info(f'''Converted in {output_file}''' )
for utils_file in utils_files:
try:
a__ : Any = os.path.basename(lowerCamelCase__ )
a__ : Optional[int] = imports_to_builder_map[f_name.replace(".py" , "" )]
self._logger.info(f'''Moving {dest_folder} to {utils_file}''' )
shutil.copy(lowerCamelCase__ , lowerCamelCase__ )
except KeyError:
self._logger.error(f'''Cannot find destination folder for {utils_file}. Please copy manually.''' )
if with_manual_update:
for file_path in with_manual_update:
self._logger.warning(
f'''You need to manually update file {file_path} to remove configurations using \'TextEncoderConfig\'.''' )
| 37 | 1 |
def UpperCamelCase_ ( __a , __a ) -> str:
if a < 0 or b < 0:
raise ValueError("the value of both inputs must be positive" )
a__ : Optional[Any] = str(bin(__a ) )[2:] # remove the leading "0b"
a__ : Optional[int] = str(bin(__a ) )[2:] # remove the leading "0b"
a__ : Optional[Any] = max(len(__a ) , len(__a ) )
return "0b" + "".join(
str(int(char_a != char_b ) )
for char_a, char_b in zip(a_binary.zfill(__a ) , b_binary.zfill(__a ) ) )
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 |
import os
from typing import Optional
import fsspec
from fsspec.archive import AbstractArchiveFileSystem
from fsspec.utils import DEFAULT_BLOCK_SIZE
class A__ ( A__ ):
"""simple docstring"""
_lowercase = ''
_lowercase = (
None # protocol passed in prefix to the url. ex: "gzip", for gzip://file.txt::http://foo.bar/file.txt.gz
)
_lowercase = None # compression type in fsspec. ex: "gzip"
_lowercase = None # extension of the filename to strip. ex: "".gz" to get file.txt from file.txt.gz
def __init__( self : List[str] , lowerCamelCase__ : str = "" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , **lowerCamelCase__ : List[str] ):
super().__init__(self , **lowerCamelCase__ )
# always open as "rb" since fsspec can then use the TextIOWrapper to make it work for "r" mode
a__ : str = fsspec.open(
lowerCamelCase__ , mode="rb" , protocol=lowerCamelCase__ , compression=self.compression , client_kwargs={
"requote_redirect_url": False, # see https://github.com/huggingface/datasets/pull/5459
"trust_env": True, # Enable reading proxy env variables.
**(target_options or {}).pop("client_kwargs" , {} ), # To avoid issues if it was already passed.
} , **(target_options or {}) , )
a__ : Optional[int] = os.path.basename(self.file.path.split("::" )[0] )
a__ : int = (
self.compressed_name[: self.compressed_name.rindex("." )]
if "." in self.compressed_name
else self.compressed_name
)
a__ : List[Any] = None
@classmethod
def _UpperCamelCase( cls : int , lowerCamelCase__ : int ):
# compressed file paths are always relative to the archive root
return super()._strip_protocol(lowerCamelCase__ ).lstrip("/" )
def _UpperCamelCase( self : Dict ):
if self.dir_cache is None:
a__ : Dict = {**self.file.fs.info(self.file.path ), "name": self.uncompressed_name}
a__ : int = {f["name"]: f}
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str ):
return self.file.open().read()
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : int=None , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[str]=None , **lowerCamelCase__ : Optional[Any] , ):
a__ : Optional[int] = self._strip_protocol(lowerCamelCase__ )
if mode != "rb":
raise ValueError(f'''Tried to read with mode {mode} on file {self.file.path} opened with mode \'rb\'''' )
return self.file.open()
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'bz2'
_lowercase = 'bz2'
_lowercase = '.bz2'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gzip'
_lowercase = 'gzip'
_lowercase = '.gz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'lz4'
_lowercase = 'lz4'
_lowercase = '.lz4'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'xz'
_lowercase = 'xz'
_lowercase = '.xz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'zstd'
_lowercase = 'zstd'
_lowercase = '.zst'
def __init__( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , lowerCamelCase__ : int = DEFAULT_BLOCK_SIZE , **lowerCamelCase__ : Tuple , ):
super().__init__(
fo=lowerCamelCase__ , mode=lowerCamelCase__ , target_protocol=lowerCamelCase__ , target_options=lowerCamelCase__ , block_size=lowerCamelCase__ , **lowerCamelCase__ , )
# We need to wrap the zstd decompressor to avoid this error in fsspec==2021.7.0 and zstandard==0.15.2:
#
# File "/Users/user/.virtualenvs/hf-datasets/lib/python3.7/site-packages/fsspec/core.py", line 145, in open
# out.close = close
# AttributeError: 'zstd.ZstdDecompressionReader' object attribute 'close' is read-only
#
# see https://github.com/intake/filesystem_spec/issues/725
a__ : Any = self.file.__enter__
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : str ):
a__ : List[Any] = file_
def __enter__( self : str ):
self._file.__enter__()
return self
def __exit__( self : int , *lowerCamelCase__ : List[str] , **lowerCamelCase__ : int ):
self._file.__exit__(*lowerCamelCase__ , **lowerCamelCase__ )
def __iter__( self : List[str] ):
return iter(self._file )
def _UpperCamelCase( self : Any ):
return next(self._file )
def __getattr__( self : Optional[Any] , lowerCamelCase__ : Tuple ):
return getattr(self._file , lowerCamelCase__ )
def fixed_enter(*lowerCamelCase__ : List[str] , **lowerCamelCase__ : str ):
return WrappedFile(_enter(*lowerCamelCase__ , **lowerCamelCase__ ) )
a__ : Any = fixed_enter
| 37 | 1 |
import os
import socket
from contextlib import contextmanager
import torch
from ..commands.config.default import write_basic_config # noqa: F401
from ..state import PartialState
from .dataclasses import DistributedType
from .imports import is_deepspeed_available, is_tpu_available
from .transformer_engine import convert_model
from .versions import is_torch_version
if is_deepspeed_available():
from deepspeed import DeepSpeedEngine
if is_tpu_available(check_device=False):
import torch_xla.core.xla_model as xm
def UpperCamelCase_ ( __a ) -> Dict:
if is_torch_version("<" , "2.0.0" ) or not hasattr(__a , "_dynamo" ):
return False
return isinstance(__a , torch._dynamo.eval_frame.OptimizedModule )
def UpperCamelCase_ ( __a , __a = True ) -> Tuple:
a__ : Union[str, Any] = (torch.nn.parallel.DistributedDataParallel, torch.nn.DataParallel)
a__ : List[Any] = is_compiled_module(__a )
if is_compiled:
a__ : Optional[int] = model
a__ : List[str] = model._orig_mod
if is_deepspeed_available():
options += (DeepSpeedEngine,)
while isinstance(__a , __a ):
a__ : int = model.module
if not keep_fpaa_wrapper:
a__ : Union[str, Any] = getattr(__a , "forward" )
a__ : Union[str, Any] = model.__dict__.pop("_original_forward" , __a )
if original_forward is not None:
while hasattr(__a , "__wrapped__" ):
a__ : int = forward.__wrapped__
if forward == original_forward:
break
a__ : Any = forward
if getattr(__a , "_converted_to_transformer_engine" , __a ):
convert_model(__a , to_transformer_engine=__a )
if is_compiled:
a__ : List[str] = model
a__ : Optional[int] = compiled_model
return model
def UpperCamelCase_ ( ) -> int:
PartialState().wait_for_everyone()
def UpperCamelCase_ ( __a , __a ) -> int:
if PartialState().distributed_type == DistributedType.TPU:
xm.save(__a , __a )
elif PartialState().local_process_index == 0:
torch.save(__a , __a )
@contextmanager
def UpperCamelCase_ ( **__a ) -> Optional[int]:
for key, value in kwargs.items():
a__ : int = str(__a )
yield
for key in kwargs:
if key.upper() in os.environ:
del os.environ[key.upper()]
def UpperCamelCase_ ( __a ) -> Dict:
if not hasattr(__a , "__qualname__" ) and not hasattr(__a , "__name__" ):
a__ : Union[str, Any] = getattr(__a , "__class__" , __a )
if hasattr(__a , "__qualname__" ):
return obj.__qualname__
if hasattr(__a , "__name__" ):
return obj.__name__
return str(__a )
def UpperCamelCase_ ( __a , __a ) -> str:
for key, value in source.items():
if isinstance(__a , __a ):
a__ : Any = destination.setdefault(__a , {} )
merge_dicts(__a , __a )
else:
a__ : List[str] = value
return destination
def UpperCamelCase_ ( __a = None ) -> bool:
if port is None:
a__ : int = 29_500
with socket.socket(socket.AF_INET , socket.SOCK_STREAM ) as s:
return s.connect_ex(("localhost", port) ) == 0
| 37 |
import argparse
import os
import numpy as np
import tensorflow as tf
import torch
from transformers import BertModel
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Union[str, Any] = ("dense.weight", "attention.self.query", "attention.self.key", "attention.self.value")
a__ : Union[str, Any] = (
("layer.", "layer_"),
("word_embeddings.weight", "word_embeddings"),
("position_embeddings.weight", "position_embeddings"),
("token_type_embeddings.weight", "token_type_embeddings"),
(".", "/"),
("LayerNorm/weight", "LayerNorm/gamma"),
("LayerNorm/bias", "LayerNorm/beta"),
("weight", "kernel"),
)
if not os.path.isdir(__a ):
os.makedirs(__a )
a__ : Any = model.state_dict()
def to_tf_var_name(__a ):
for patt, repl in iter(__a ):
a__ : Tuple = name.replace(__a , __a )
return f'''bert/{name}'''
def create_tf_var(__a , __a , __a ):
a__ : Tuple = tf.dtypes.as_dtype(tensor.dtype )
a__ : Dict = tf.get_variable(dtype=__a , shape=tensor.shape , name=__a , initializer=tf.zeros_initializer() )
session.run(tf.variables_initializer([tf_var] ) )
session.run(__a )
return tf_var
tf.reset_default_graph()
with tf.Session() as session:
for var_name in state_dict:
a__ : int = to_tf_var_name(__a )
a__ : Union[str, Any] = state_dict[var_name].numpy()
if any(x in var_name for x in tensors_to_transpose ):
a__ : int = torch_tensor.T
a__ : Optional[Any] = create_tf_var(tensor=__a , name=__a , session=__a )
tf.keras.backend.set_value(__a , __a )
a__ : int = session.run(__a )
print(f'''Successfully created {tf_name}: {np.allclose(__a , __a )}''' )
a__ : Any = tf.train.Saver(tf.trainable_variables() )
saver.save(__a , os.path.join(__a , model_name.replace("-" , "_" ) + ".ckpt" ) )
def UpperCamelCase_ ( __a=None ) -> int:
a__ : Dict = argparse.ArgumentParser()
parser.add_argument("--model_name" , type=__a , required=__a , help="model name e.g. bert-base-uncased" )
parser.add_argument(
"--cache_dir" , type=__a , default=__a , required=__a , help="Directory containing pytorch model" )
parser.add_argument("--pytorch_model_path" , type=__a , required=__a , help="/path/to/<pytorch-model-name>.bin" )
parser.add_argument("--tf_cache_dir" , type=__a , required=__a , help="Directory in which to save tensorflow model" )
a__ : Optional[Any] = parser.parse_args(__a )
a__ : Tuple = BertModel.from_pretrained(
pretrained_model_name_or_path=args.model_name , state_dict=torch.load(args.pytorch_model_path ) , cache_dir=args.cache_dir , )
convert_pytorch_checkpoint_to_tf(model=__a , ckpt_dir=args.tf_cache_dir , model_name=args.model_name )
if __name__ == "__main__":
main()
| 37 | 1 |
from ...configuration_utils import PretrainedConfig
from ...utils import logging
UpperCamelCase : Dict = logging.get_logger(__name__)
UpperCamelCase : Tuple = {
"""naver-clova-ix/donut-base""": """https://huggingface.co/naver-clova-ix/donut-base/resolve/main/config.json""",
# See all Donut models at https://huggingface.co/models?filter=donut-swin
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'donut-swin'
_lowercase = {
'num_attention_heads': 'num_heads',
'num_hidden_layers': 'num_layers',
}
def __init__( self : Any , lowerCamelCase__ : Optional[Any]=224 , lowerCamelCase__ : Tuple=4 , lowerCamelCase__ : str=3 , lowerCamelCase__ : Union[str, Any]=96 , lowerCamelCase__ : Optional[int]=[2, 2, 6, 2] , lowerCamelCase__ : str=[3, 6, 12, 24] , lowerCamelCase__ : List[str]=7 , lowerCamelCase__ : int=4.0 , lowerCamelCase__ : Any=True , lowerCamelCase__ : Tuple=0.0 , lowerCamelCase__ : Optional[Any]=0.0 , lowerCamelCase__ : int=0.1 , lowerCamelCase__ : List[Any]="gelu" , lowerCamelCase__ : List[str]=False , lowerCamelCase__ : Optional[int]=0.02 , lowerCamelCase__ : Dict=1E-5 , **lowerCamelCase__ : Optional[int] , ):
super().__init__(**lowerCamelCase__ )
a__ : Tuple = image_size
a__ : str = patch_size
a__ : Tuple = num_channels
a__ : Any = embed_dim
a__ : int = depths
a__ : List[str] = len(lowerCamelCase__ )
a__ : Optional[Any] = num_heads
a__ : Dict = window_size
a__ : Optional[int] = mlp_ratio
a__ : List[Any] = qkv_bias
a__ : Tuple = hidden_dropout_prob
a__ : int = attention_probs_dropout_prob
a__ : Union[str, Any] = drop_path_rate
a__ : List[str] = hidden_act
a__ : Optional[Any] = use_absolute_embeddings
a__ : Union[str, Any] = layer_norm_eps
a__ : Optional[int] = initializer_range
# we set the hidden_size attribute in order to make Swin work with VisionEncoderDecoderModel
# this indicates the channel dimension after the last stage of the model
a__ : Any = int(embed_dim * 2 ** (len(lowerCamelCase__ ) - 1) )
| 37 |
import inspect
import unittest
from huggingface_hub import hf_hub_download
from transformers import ASTConfig
from transformers.testing_utils import require_torch, require_torchaudio, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_torchaudio_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import ASTForAudioClassification, ASTModel
from transformers.models.audio_spectrogram_transformer.modeling_audio_spectrogram_transformer import (
AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
)
if is_torchaudio_available():
import torchaudio
from transformers import ASTFeatureExtractor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[str] , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[Any]=2 , lowerCamelCase__ : Any=24 , lowerCamelCase__ : Optional[Any]=16 , lowerCamelCase__ : int=True , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[Any]=32 , lowerCamelCase__ : List[str]=5 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Optional[Any]=37 , lowerCamelCase__ : Any="gelu" , lowerCamelCase__ : Union[str, Any]=0.1 , lowerCamelCase__ : Optional[int]=0.1 , lowerCamelCase__ : str=10 , lowerCamelCase__ : Optional[Any]=0.02 , lowerCamelCase__ : str=None , lowerCamelCase__ : List[str]=2 , lowerCamelCase__ : Optional[Any]=2 , ):
a__ : str = parent
a__ : Any = batch_size
a__ : Dict = patch_size
a__ : List[Any] = max_length
a__ : str = num_mel_bins
a__ : Optional[Any] = is_training
a__ : Optional[int] = use_labels
a__ : List[Any] = hidden_size
a__ : str = num_hidden_layers
a__ : Any = num_attention_heads
a__ : Union[str, Any] = intermediate_size
a__ : List[str] = hidden_act
a__ : str = hidden_dropout_prob
a__ : Tuple = attention_probs_dropout_prob
a__ : List[Any] = type_sequence_label_size
a__ : Any = initializer_range
a__ : str = scope
a__ : List[str] = frequency_stride
a__ : Union[str, Any] = time_stride
# in AST, the seq length equals the number of patches + 2 (we add 2 for the [CLS] and distillation tokens)
a__ : List[Any] = (self.num_mel_bins - self.patch_size) // self.frequency_stride + 1
a__ : List[str] = (self.max_length - self.patch_size) // self.time_stride + 1
a__ : Tuple = frequency_out_dimension * time_out_dimension
a__ : List[str] = num_patches + 2
def _UpperCamelCase( self : List[str] ):
a__ : Any = floats_tensor([self.batch_size, self.max_length, self.num_mel_bins] )
a__ : List[Any] = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : List[str] = self.get_config()
return config, input_values, labels
def _UpperCamelCase( self : Optional[int] ):
return ASTConfig(
patch_size=self.patch_size , max_length=self.max_length , num_mel_bins=self.num_mel_bins , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , frequency_stride=self.frequency_stride , time_stride=self.time_stride , )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : int , lowerCamelCase__ : Optional[int] ):
a__ : List[Any] = ASTModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Dict = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : str ):
a__ : Dict = self.prepare_config_and_inputs()
(
(
a__
), (
a__
), (
a__
),
) : Optional[int] = config_and_inputs
a__ : List[Any] = {"input_values": input_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(
ASTModel,
ASTForAudioClassification,
)
if is_torch_available()
else ()
)
_lowercase = (
{'audio-classification': ASTForAudioClassification, 'feature-extraction': ASTModel}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
if pipeline_test_casse_name == "AudioClassificationPipelineTests":
return True
return False
def _UpperCamelCase( self : str ):
a__ : str = ASTModelTester(self )
a__ : Any = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[str] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="AST does not use inputs_embeds" )
def _UpperCamelCase( self : List[str] ):
pass
def _UpperCamelCase( self : Optional[int] ):
a__, a__ : str = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Any = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Union[str, Any] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : Tuple ):
a__, a__ : List[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Dict = model_class(lowerCamelCase__ )
a__ : Optional[int] = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Optional[Any] = ["input_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
for model_name in AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Union[str, Any] = ASTModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : Optional[int] = hf_hub_download(
repo_id="nielsr/audio-spectogram-transformer-checkpoint" , filename="sample_audio.flac" , repo_type="dataset" )
a__, a__ : List[str] = torchaudio.load(__a )
return audio, sampling_rate
@require_torch
@require_torchaudio
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : List[str] ):
return (
ASTFeatureExtractor.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" )
if is_torchaudio_available()
else None
)
@slow
def _UpperCamelCase( self : Optional[int] ):
a__ : int = self.default_feature_extractor
a__ : Optional[Any] = ASTForAudioClassification.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" ).to(lowerCamelCase__ )
a__ : Any = self.default_feature_extractor
a__, a__ : Dict = prepare_audio()
a__ : str = audio.squeeze().numpy()
a__ : Any = feature_extractor(lowerCamelCase__ , sampling_rate=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(**lowerCamelCase__ )
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 527) )
self.assertEqual(outputs.logits.shape , lowerCamelCase__ )
a__ : List[str] = torch.tensor([-0.8760, -7.0042, -8.6602] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(outputs.logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
| 37 | 1 |
import json
import os
from functools import lru_cache
from typing import List, Optional, Tuple
import regex as re
from ...tokenization_utils import AddedToken, PreTrainedTokenizer
from ...utils import logging
UpperCamelCase : Union[str, Any] = logging.get_logger(__name__)
UpperCamelCase : int = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt"""}
# See all BART models at https://huggingface.co/models?filter=bart
UpperCamelCase : str = {
"""vocab_file""": {
"""facebook/bart-base""": """https://huggingface.co/facebook/bart-base/resolve/main/vocab.json""",
"""facebook/bart-large""": """https://huggingface.co/facebook/bart-large/resolve/main/vocab.json""",
"""facebook/bart-large-mnli""": """https://huggingface.co/facebook/bart-large-mnli/resolve/main/vocab.json""",
"""facebook/bart-large-cnn""": """https://huggingface.co/facebook/bart-large-cnn/resolve/main/vocab.json""",
"""facebook/bart-large-xsum""": """https://huggingface.co/facebook/bart-large-xsum/resolve/main/vocab.json""",
"""yjernite/bart_eli5""": """https://huggingface.co/yjernite/bart_eli5/resolve/main/vocab.json""",
},
"""merges_file""": {
"""facebook/bart-base""": """https://huggingface.co/facebook/bart-base/resolve/main/merges.txt""",
"""facebook/bart-large""": """https://huggingface.co/facebook/bart-large/resolve/main/merges.txt""",
"""facebook/bart-large-mnli""": """https://huggingface.co/facebook/bart-large-mnli/resolve/main/merges.txt""",
"""facebook/bart-large-cnn""": """https://huggingface.co/facebook/bart-large-cnn/resolve/main/merges.txt""",
"""facebook/bart-large-xsum""": """https://huggingface.co/facebook/bart-large-xsum/resolve/main/merges.txt""",
"""yjernite/bart_eli5""": """https://huggingface.co/yjernite/bart_eli5/resolve/main/merges.txt""",
},
}
UpperCamelCase : Any = {
"""facebook/bart-base""": 1024,
"""facebook/bart-large""": 1024,
"""facebook/bart-large-mnli""": 1024,
"""facebook/bart-large-cnn""": 1024,
"""facebook/bart-large-xsum""": 1024,
"""yjernite/bart_eli5""": 1024,
}
@lru_cache()
def UpperCamelCase_ ( ) -> Dict:
a__ : Any = (
list(range(ord("!" ) , ord("~" ) + 1 ) ) + list(range(ord("¡" ) , ord("¬" ) + 1 ) ) + list(range(ord("®" ) , ord("ÿ" ) + 1 ) )
)
a__ : Tuple = bs[:]
a__ : Any = 0
for b in range(2**8 ):
if b not in bs:
bs.append(__a )
cs.append(2**8 + n )
n += 1
a__ : str = [chr(__a ) for n in cs]
return dict(zip(__a , __a ) )
def UpperCamelCase_ ( __a ) -> Dict:
a__ : int = set()
a__ : int = word[0]
for char in word[1:]:
pairs.add((prev_char, char) )
a__ : str = char
return pairs
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = ['input_ids', 'attention_mask']
def __init__( self : Optional[int] , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int="replace" , lowerCamelCase__ : Dict="<s>" , lowerCamelCase__ : str="</s>" , lowerCamelCase__ : Any="</s>" , lowerCamelCase__ : Optional[Any]="<s>" , lowerCamelCase__ : str="<unk>" , lowerCamelCase__ : Optional[int]="<pad>" , lowerCamelCase__ : Dict="<mask>" , lowerCamelCase__ : Any=False , **lowerCamelCase__ : Optional[int] , ):
a__ : Tuple = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else bos_token
a__ : Tuple = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else eos_token
a__ : List[str] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else sep_token
a__ : List[str] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else cls_token
a__ : Optional[int] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else unk_token
a__ : Any = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else pad_token
# Mask token behave like a normal word, i.e. include the space before it
a__ : Dict = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else mask_token
super().__init__(
errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , **lowerCamelCase__ , )
with open(lowerCamelCase__ , encoding="utf-8" ) as vocab_handle:
a__ : int = json.load(lowerCamelCase__ )
a__ : Optional[int] = {v: k for k, v in self.encoder.items()}
a__ : int = errors # how to handle errors in decoding
a__ : Union[str, Any] = bytes_to_unicode()
a__ : int = {v: k for k, v in self.byte_encoder.items()}
with open(lowerCamelCase__ , encoding="utf-8" ) as merges_handle:
a__ : str = merges_handle.read().split("\n" )[1:-1]
a__ : Dict = [tuple(merge.split() ) for merge in bpe_merges]
a__ : List[str] = dict(zip(lowerCamelCase__ , range(len(lowerCamelCase__ ) ) ) )
a__ : Optional[int] = {}
a__ : Tuple = add_prefix_space
# Should have added re.IGNORECASE so BPE merges can happen for capitalized versions of contractions
a__ : List[str] = re.compile(r"'s|'t|'re|'ve|'m|'ll|'d| ?\p{L}+| ?\p{N}+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+" )
@property
def _UpperCamelCase( self : Union[str, Any] ):
return len(self.encoder )
def _UpperCamelCase( self : int ):
return dict(self.encoder , **self.added_tokens_encoder )
def _UpperCamelCase( self : Any , lowerCamelCase__ : Dict ):
if token in self.cache:
return self.cache[token]
a__ : Optional[Any] = tuple(lowerCamelCase__ )
a__ : Optional[int] = get_pairs(lowerCamelCase__ )
if not pairs:
return token
while True:
a__ : List[str] = min(lowerCamelCase__ , key=lambda lowerCamelCase__ : self.bpe_ranks.get(lowerCamelCase__ , float("inf" ) ) )
if bigram not in self.bpe_ranks:
break
a__, a__ : Union[str, Any] = bigram
a__ : Any = []
a__ : Union[str, Any] = 0
while i < len(lowerCamelCase__ ):
try:
a__ : List[str] = word.index(lowerCamelCase__ , lowerCamelCase__ )
except ValueError:
new_word.extend(word[i:] )
break
else:
new_word.extend(word[i:j] )
a__ : Optional[int] = j
if word[i] == first and i < len(lowerCamelCase__ ) - 1 and word[i + 1] == second:
new_word.append(first + second )
i += 2
else:
new_word.append(word[i] )
i += 1
a__ : Dict = tuple(lowerCamelCase__ )
a__ : int = new_word
if len(lowerCamelCase__ ) == 1:
break
else:
a__ : Optional[int] = get_pairs(lowerCamelCase__ )
a__ : List[str] = " ".join(lowerCamelCase__ )
a__ : List[Any] = word
return word
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Tuple ):
a__ : Optional[int] = []
for token in re.findall(self.pat , lowerCamelCase__ ):
a__ : str = "".join(
self.byte_encoder[b] for b in token.encode("utf-8" ) ) # Maps all our bytes to unicode strings, avoiding control tokens of the BPE (spaces in our case)
bpe_tokens.extend(bpe_token for bpe_token in self.bpe(lowerCamelCase__ ).split(" " ) )
return bpe_tokens
def _UpperCamelCase( self : str , lowerCamelCase__ : Union[str, Any] ):
return self.encoder.get(lowerCamelCase__ , self.encoder.get(self.unk_token ) )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Optional[int] ):
return self.decoder.get(lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Dict ):
a__ : List[Any] = "".join(lowerCamelCase__ )
a__ : Optional[int] = bytearray([self.byte_decoder[c] for c in text] ).decode("utf-8" , errors=self.errors )
return text
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
if not os.path.isdir(lowerCamelCase__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
a__ : str = os.path.join(
lowerCamelCase__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] )
a__ : Optional[int] = os.path.join(
lowerCamelCase__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["merges_file"] )
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as f:
f.write(json.dumps(self.encoder , indent=2 , sort_keys=lowerCamelCase__ , ensure_ascii=lowerCamelCase__ ) + "\n" )
a__ : int = 0
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as writer:
writer.write("#version: 0.2\n" )
for bpe_tokens, token_index in sorted(self.bpe_ranks.items() , key=lambda lowerCamelCase__ : kv[1] ):
if index != token_index:
logger.warning(
f'''Saving vocabulary to {merge_file}: BPE merge indices are not consecutive.'''
" Please check that the tokenizer is not corrupted!" )
a__ : List[str] = token_index
writer.write(" ".join(lowerCamelCase__ ) + "\n" )
index += 1
return vocab_file, merge_file
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
if token_ids_a is None:
return [self.cls_token_id] + token_ids_a + [self.sep_token_id]
a__ : Any = [self.cls_token_id]
a__ : str = [self.sep_token_id]
return cls + token_ids_a + sep + sep + token_ids_a + sep
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None , lowerCamelCase__ : bool = False ):
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=lowerCamelCase__ , token_ids_a=lowerCamelCase__ , already_has_special_tokens=lowerCamelCase__ )
if token_ids_a is None:
return [1] + ([0] * len(lowerCamelCase__ )) + [1]
return [1] + ([0] * len(lowerCamelCase__ )) + [1, 1] + ([0] * len(lowerCamelCase__ )) + [1]
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : int = [self.sep_token_id]
a__ : List[str] = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
def _UpperCamelCase( self : str , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Any=False , **lowerCamelCase__ : Union[str, Any] ):
a__ : int = kwargs.pop("add_prefix_space" , self.add_prefix_space )
if (is_split_into_words or add_prefix_space) and (len(lowerCamelCase__ ) > 0 and not text[0].isspace()):
a__ : Any = " " + text
return (text, kwargs)
| 37 |
import pickle
import shutil
import tempfile
import unittest
from transformers import SPIECE_UNDERLINE, XGLMTokenizer, XGLMTokenizerFast
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers, slow
from transformers.utils import cached_property
from ...test_tokenization_common import TokenizerTesterMixin
UpperCamelCase : Optional[Any] = get_tests_dir("""fixtures/test_sentencepiece.model""")
@require_sentencepiece
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = XGLMTokenizer
_lowercase = XGLMTokenizerFast
_lowercase = True
_lowercase = True
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# We have a SentencePiece fixture for testing
a__ : str = XGLMTokenizer(lowerCamelCase__ , keep_accents=lowerCamelCase__ )
tokenizer.save_pretrained(self.tmpdirname )
def _UpperCamelCase( self : List[Any] ):
a__ : int = "<pad>"
a__ : Union[str, Any] = 1
self.assertEqual(self.get_tokenizer()._convert_token_to_id(lowerCamelCase__ ) , lowerCamelCase__ )
self.assertEqual(self.get_tokenizer()._convert_id_to_token(lowerCamelCase__ ) , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : List[str] = list(self.get_tokenizer().get_vocab().keys() )
self.assertEqual(vocab_keys[0] , "<s>" )
self.assertEqual(vocab_keys[1] , "<pad>" )
self.assertEqual(len(lowerCamelCase__ ) , 1_008 )
def _UpperCamelCase( self : Dict ):
self.assertEqual(self.get_tokenizer().vocab_size , 1_008 )
def _UpperCamelCase( self : Optional[int] ):
a__ : str = XGLMTokenizer(lowerCamelCase__ , keep_accents=lowerCamelCase__ )
a__ : List[str] = tokenizer.tokenize("This is a test" )
self.assertListEqual(lowerCamelCase__ , ["▁This", "▁is", "▁a", "▁t", "est"] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
a__ : Any = tokenizer.tokenize("I was born in 92000, and this is falsé." )
self.assertListEqual(
lowerCamelCase__ , [
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"9",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"é",
".",
] , )
a__ : List[str] = tokenizer.convert_tokens_to_ids(lowerCamelCase__ )
self.assertListEqual(
lowerCamelCase__ , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
a__ : Dict = tokenizer.convert_ids_to_tokens(lowerCamelCase__ )
self.assertListEqual(
lowerCamelCase__ , [
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"<unk>",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"<unk>",
".",
] , )
@cached_property
def _UpperCamelCase( self : Dict ):
return XGLMTokenizer.from_pretrained("facebook/xglm-564M" )
def _UpperCamelCase( self : Union[str, Any] ):
with tempfile.NamedTemporaryFile() as f:
shutil.copyfile(lowerCamelCase__ , f.name )
a__ : Any = XGLMTokenizer(f.name , keep_accents=lowerCamelCase__ )
a__ : List[str] = pickle.dumps(lowerCamelCase__ )
pickle.loads(lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
if not self.test_rust_tokenizer:
return
a__ : Any = self.get_tokenizer()
a__ : Optional[Any] = self.get_rust_tokenizer()
a__ : Tuple = "I was born in 92000, and this is falsé."
a__ : List[str] = tokenizer.tokenize(lowerCamelCase__ )
a__ : Union[str, Any] = rust_tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : Optional[int] = tokenizer.encode(lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
a__ : Union[str, Any] = rust_tokenizer.encode(lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : List[str] = self.get_rust_tokenizer()
a__ : Tuple = tokenizer.encode(lowerCamelCase__ )
a__ : Optional[Any] = rust_tokenizer.encode(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = "Hello World!"
a__ : List[str] = [2, 31_227, 4_447, 35]
self.assertListEqual(lowerCamelCase__ , self.big_tokenizer.encode(lowerCamelCase__ ) )
@slow
def _UpperCamelCase( self : Union[str, Any] ):
a__ : Optional[int] = (
"This is a very long text with a lot of weird characters, such as: . , ~ ? ( ) \" [ ] ! : - . Also we will"
" add words that should not exsist and be tokenized to unk, such as saoneuhaoesuth"
)
# fmt: off
a__ : Union[str, Any] = [2, 1_018, 67, 11, 1_988, 2_617, 5_631, 278, 11, 3_407, 48, 71_630, 28_085, 4, 3_234, 157, 13, 6, 5, 6, 4, 3_526, 768, 15, 659, 57, 298, 3_983, 864, 129, 21, 6, 5, 13_675, 377, 652, 7_580, 10_341, 155, 2_817, 422, 1_666, 7, 1_674, 53, 113, 202_277, 17_892, 33, 60, 87, 4, 3_234, 157, 61, 2_667, 52_376, 19, 88, 23, 735]
# fmt: on
self.assertListEqual(lowerCamelCase__ , self.big_tokenizer.encode(lowerCamelCase__ ) )
@slow
def _UpperCamelCase( self : List[Any] ):
# fmt: off
a__ : Optional[int] = {
"input_ids": [[2, 108_825, 1_163, 15, 88_010, 473, 15_898, 157, 13_672, 1_857, 312, 8, 238_021, 1_163, 53, 13_672, 1_857, 312, 8, 53_283, 182_396, 8, 18_566, 16, 36_733, 4_101, 8, 230, 244_017, 122_553, 7, 15, 132_597, 4, 293, 12_511, 7_610, 4, 3_414, 132_597, 9, 4, 32_361, 362, 4, 734, 28_512, 32_569, 18, 4, 32_361, 26_096, 14_982, 73, 18_715, 21_433, 235_261, 15, 492, 12_427, 16, 53, 18_715, 21_433, 65_454, 15, 23_659, 563, 16, 278, 597, 2_843, 595, 7_931, 182_396, 64_186, 22, 886, 595, 132_981, 53, 25_540, 3_449, 43_982, 39_901, 5_951, 878, 330, 4, 27_694, 80_269, 312, 53, 6_517, 11_780, 611, 20_408, 5], [2, 6, 132_597, 67, 42_897, 33, 592, 8, 163_729, 25_540, 361, 136_997, 109_514, 173_230, 7, 501, 60, 102_913, 196, 5_631, 235, 63_243, 473, 6, 231_757, 74, 5_277, 7_905, 53, 3_095, 37_317, 22, 454, 183_874, 5], [2, 268, 31_298, 46_530, 6, 132_935, 43_831, 7, 597, 32, 24, 3_688, 9_865, 5]],
"attention_mask": [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]
} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=lowerCamelCase__ , model_name="facebook/xglm-564M" , padding=lowerCamelCase__ , )
| 37 | 1 |
import logging
import os
import random
import sys
from dataclasses import dataclass, field
from typing import Optional
import datasets
import numpy as np
import pandas as pd
from datasets import load_dataset
import transformers
from transformers import (
AutoConfig,
BartForSequenceClassification,
DataCollatorWithPadding,
EvalPrediction,
HfArgumentParser,
TapexTokenizer,
Trainer,
TrainingArguments,
default_data_collator,
set_seed,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("""4.17.0.dev0""")
require_version("""datasets>=1.8.0""", """To fix: pip install -r examples/pytorch/text-classification/requirements.txt""")
UpperCamelCase : str = logging.getLogger(__name__)
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(
default='tab_fact' , metadata={'help': 'The name of the dataset to use (via the datasets library).'} )
_lowercase = field(
default='tab_fact' , metadata={'help': 'The configuration name of the dataset to use (via the datasets library).'} , )
_lowercase = field(
default=1_0_2_4 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
_lowercase = field(
default=A__ , metadata={'help': 'Overwrite the cached preprocessed datasets or not.'} )
_lowercase = field(
default=A__ , metadata={
'help': (
'Whether to pad all samples to `max_seq_length`. '
'If False, will pad the samples dynamically when batching to the maximum length in the batch.'
)
} , )
_lowercase = field(
default=A__ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of training examples to this '
'value if set.'
)
} , )
_lowercase = field(
default=A__ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of evaluation examples to this '
'value if set.'
)
} , )
_lowercase = field(
default=A__ , metadata={
'help': (
'For debugging purposes or quicker training, truncate the number of prediction examples to this '
'value if set.'
)
} , )
_lowercase = field(
default=A__ , metadata={'help': 'A csv or a json file containing the training data.'} )
_lowercase = field(
default=A__ , metadata={'help': 'A csv or a json file containing the validation data.'} )
_lowercase = field(default=A__ , metadata={'help': 'A csv or a json file containing the test data.'} )
def _UpperCamelCase( self : int ):
if self.dataset_name is not None:
pass
elif self.train_file is None or self.validation_file is None:
raise ValueError("Need either a GLUE task, a training/validation file or a dataset name." )
else:
a__ : Dict = self.train_file.split("." )[-1]
assert train_extension in ["csv", "json"], "`train_file` should be a csv or a json file."
a__ : int = self.validation_file.split("." )[-1]
assert (
validation_extension == train_extension
), "`validation_file` should have the same extension (csv or json) as `train_file`."
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(
default=A__ , metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
_lowercase = field(
default=A__ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
_lowercase = field(
default=A__ , metadata={'help': 'Whether to use one of the fast tokenizer (backed by the tokenizers library) or not.'} , )
_lowercase = field(
default='main' , metadata={'help': 'The specific model version to use (can be a branch name, tag name or commit id).'} , )
_lowercase = field(
default=A__ , metadata={
'help': (
'Will use the token generated when running `huggingface-cli login` (necessary to use this script '
'with private models).'
)
} , )
def UpperCamelCase_ ( ) -> Dict:
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
a__ : Union[str, Any] = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments) )
if len(sys.argv ) == 2 and sys.argv[1].endswith(".json" ):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
a__, a__, a__ : List[str] = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1] ) )
else:
a__, a__, a__ : Optional[Any] = parser.parse_args_into_dataclasses()
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s" , datefmt="%m/%d/%Y %H:%M:%S" , handlers=[logging.StreamHandler(sys.stdout )] , )
a__ : Tuple = training_args.get_process_log_level()
logger.setLevel(__a )
datasets.utils.logging.set_verbosity(__a )
transformers.utils.logging.set_verbosity(__a )
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Log on each process the small summary:
logger.warning(
f'''Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}'''
+ f'''distributed training: {bool(training_args.local_rank != -1 )}, 16-bits training: {training_args.fpaa}''' )
logger.info(f'''Training/evaluation parameters {training_args}''' )
# Detecting last checkpoint.
a__ : Optional[int] = None
if os.path.isdir(training_args.output_dir ) and training_args.do_train and not training_args.overwrite_output_dir:
a__ : int = get_last_checkpoint(training_args.output_dir )
if last_checkpoint is None and len(os.listdir(training_args.output_dir ) ) > 0:
raise ValueError(
f'''Output directory ({training_args.output_dir}) already exists and is not empty. '''
"Use --overwrite_output_dir to overcome." )
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f'''Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change '''
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch." )
# Set seed before initializing model.
set_seed(training_args.seed )
# Get the datasets: you can either provide your own CSV/JSON training and evaluation files (see below)
# or specify a GLUE benchmark task (the dataset will be downloaded automatically from the datasets Hub).
#
# For JSON files, this script will use the `question` column for the input question and `table` column for the corresponding table.
#
# If the CSVs/JSONs contain only one non-label column, the script does single sentence classification on this
# single column. You can easily tweak this behavior (see below)
#
# In distributed training, the load_dataset function guarantee that only one local process can concurrently
# download the dataset.
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
a__ : Union[str, Any] = load_dataset(
data_args.dataset_name , data_args.dataset_config_name , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from your local files.
# CSV/JSON training and evaluation files are needed.
a__ : str = {"train": data_args.train_file, "validation": data_args.validation_file}
# Get the test dataset: you can provide your own CSV/JSON test file (see below)
# when you use `do_predict` without specifying a GLUE benchmark task.
if training_args.do_predict:
if data_args.test_file is not None:
a__ : Any = data_args.train_file.split("." )[-1]
a__ : Dict = data_args.test_file.split("." )[-1]
assert (
test_extension == train_extension
), "`test_file` should have the same extension (csv or json) as `train_file`."
a__ : Tuple = data_args.test_file
else:
raise ValueError("Need either a GLUE task or a test file for `do_predict`." )
for key in data_files.keys():
logger.info(f'''load a local file for {key}: {data_files[key]}''' )
if data_args.train_file.endswith(".csv" ):
# Loading a dataset from local csv files
a__ : List[str] = load_dataset("csv" , data_files=__a , cache_dir=model_args.cache_dir )
else:
# Loading a dataset from local json files
a__ : Tuple = load_dataset("json" , data_files=__a , cache_dir=model_args.cache_dir )
# See more about loading any type of standard or custom dataset at
# https://huggingface.co/docs/datasets/loading_datasets.html.
# Labels
a__ : Dict = raw_datasets["train"].features["label"].names
a__ : Optional[Any] = len(__a )
# Load pretrained model and tokenizer
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
a__ : Any = AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=__a , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# load tapex tokenizer
a__ : Union[str, Any] = TapexTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , use_fast=model_args.use_fast_tokenizer , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , add_prefix_space=__a , )
a__ : Any = BartForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_tf=bool(".ckpt" in model_args.model_name_or_path ) , config=__a , cache_dir=model_args.cache_dir , revision=model_args.model_revision , use_auth_token=True if model_args.use_auth_token else None , )
# Padding strategy
if data_args.pad_to_max_length:
a__ : Optional[Any] = "max_length"
else:
# We will pad later, dynamically at batch creation, to the max sequence length in each batch
a__ : Optional[Any] = False
# Some models have set the order of the labels to use, so let's make sure we do use it.
a__ : Dict = {"Refused": 0, "Entailed": 1}
a__ : List[Any] = {0: "Refused", 1: "Entailed"}
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warning(
f'''The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the'''
f'''model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}.''' )
a__ : Union[str, Any] = min(data_args.max_seq_length , tokenizer.model_max_length )
def preprocess_tabfact_function(__a ):
# Tokenize the texts
def _convert_table_text_to_pandas(__a ):
a__ : List[str] = [_table_row.split("#" ) for _table_row in _table_text.strip("\n" ).split("\n" )]
a__ : List[Any] = pd.DataFrame.from_records(_table_content[1:] , columns=_table_content[0] )
return _table_pd
a__ : str = examples["statement"]
a__ : Tuple = list(map(_convert_table_text_to_pandas , examples["table_text"] ) )
a__ : str = tokenizer(__a , __a , padding=__a , max_length=__a , truncation=__a )
a__ : Tuple = examples["label"]
return result
with training_args.main_process_first(desc="dataset map pre-processing" ):
a__ : str = raw_datasets.map(
__a , batched=__a , load_from_cache_file=not data_args.overwrite_cache , desc="Running tokenizer on dataset" , )
if training_args.do_train:
if "train" not in raw_datasets:
raise ValueError("--do_train requires a train dataset" )
a__ : Optional[Any] = raw_datasets["train"]
if data_args.max_train_samples is not None:
a__ : Optional[Any] = train_dataset.select(range(data_args.max_train_samples ) )
if training_args.do_eval:
if "validation" not in raw_datasets and "validation_matched" not in raw_datasets:
raise ValueError("--do_eval requires a validation dataset" )
a__ : Dict = raw_datasets["validation"]
if data_args.max_eval_samples is not None:
a__ : int = eval_dataset.select(range(data_args.max_eval_samples ) )
if training_args.do_predict or data_args.test_file is not None:
if "test" not in raw_datasets and "test_matched" not in raw_datasets:
raise ValueError("--do_predict requires a test dataset" )
a__ : Dict = raw_datasets["test"]
if data_args.max_predict_samples is not None:
a__ : Union[str, Any] = predict_dataset.select(range(data_args.max_predict_samples ) )
# Log a few random samples from the training set:
if training_args.do_train:
for index in random.sample(range(len(__a ) ) , 3 ):
logger.info(f'''Sample {index} of the training set: {train_dataset[index]}.''' )
# You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
# predictions and label_ids field) and has to return a dictionary string to float.
def compute_metrics(__a ):
a__ : Dict = p.predictions[0] if isinstance(p.predictions , __a ) else p.predictions
a__ : Optional[Any] = np.argmax(__a , axis=1 )
return {"accuracy": (preds == p.label_ids).astype(np.floataa ).mean().item()}
# Data collator will default to DataCollatorWithPadding, so we change it if we already did the padding.
if data_args.pad_to_max_length:
a__ : Optional[int] = default_data_collator
elif training_args.fpaa:
a__ : int = DataCollatorWithPadding(__a , pad_to_multiple_of=8 )
else:
a__ : Any = None
# Initialize our Trainer
a__ : Optional[Any] = Trainer(
model=__a , args=__a , train_dataset=train_dataset if training_args.do_train else None , eval_dataset=eval_dataset if training_args.do_eval else None , compute_metrics=__a , tokenizer=__a , data_collator=__a , )
# Training
if training_args.do_train:
a__ : Union[str, Any] = None
if training_args.resume_from_checkpoint is not None:
a__ : Dict = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
a__ : Tuple = last_checkpoint
a__ : Union[str, Any] = trainer.train(resume_from_checkpoint=__a )
a__ : int = train_result.metrics
a__ : Union[str, Any] = (
data_args.max_train_samples if data_args.max_train_samples is not None else len(__a )
)
a__ : Optional[int] = min(__a , len(__a ) )
trainer.save_model() # Saves the tokenizer too for easy upload
trainer.log_metrics("train" , __a )
trainer.save_metrics("train" , __a )
trainer.save_state()
# Evaluation
if training_args.do_eval:
logger.info("*** Evaluate ***" )
a__ : Any = trainer.evaluate(eval_dataset=__a )
a__ : Optional[Any] = data_args.max_eval_samples if data_args.max_eval_samples is not None else len(__a )
a__ : Optional[int] = min(__a , len(__a ) )
trainer.log_metrics("eval" , __a )
trainer.save_metrics("eval" , __a )
if training_args.do_predict:
logger.info("*** Predict ***" )
# Removing the `label` columns because it contains -1 and Trainer won't like that.
a__ : Optional[Any] = predict_dataset.remove_columns("label" )
a__ : List[str] = trainer.predict(__a , metric_key_prefix="predict" ).predictions
a__ : Dict = np.argmax(__a , axis=1 )
a__ : Union[str, Any] = os.path.join(training_args.output_dir , "predict_results_tabfact.txt" )
if trainer.is_world_process_zero():
with open(__a , "w" ) as writer:
logger.info("***** Predict Results *****" )
writer.write("index\tprediction\n" )
for index, item in enumerate(__a ):
a__ : Union[str, Any] = label_list[item]
writer.write(f'''{index}\t{item}\n''' )
a__ : Any = {"finetuned_from": model_args.model_name_or_path, "tasks": "text-classification"}
if training_args.push_to_hub:
trainer.push_to_hub(**__a )
else:
trainer.create_model_card(**__a )
def UpperCamelCase_ ( __a ) -> int:
# For xla_spawn (TPUs)
main()
if __name__ == "__main__":
main()
| 37 |
import argparse
import requests
import torch
# pip3 install salesforce-lavis
# I'm actually installing a slightly modified version: pip3 install git+https://github.com/nielsrogge/LAVIS.git@fix_lavis_float32 (there's also the fix_lavis branch)
# also note: to convert Vicuna checkpoints, we had to include /home/niels/python_projects/checkpoints/FastChat/vicuna-7b in lavis/configs/models/blip2/blip2_instruct_vicuna7b.yaml
# same for Vicuna-13b
from lavis.models import load_model_and_preprocess
from PIL import Image
from transformers import (
AutoTokenizer,
BlipImageProcessor,
InstructBlipConfig,
InstructBlipForConditionalGeneration,
InstructBlipProcessor,
InstructBlipQFormerConfig,
InstructBlipVisionConfig,
LlamaConfig,
LlamaTokenizerFast,
TaConfig,
TaTokenizerFast,
)
from transformers.utils.constants import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD
def UpperCamelCase_ ( ) -> int:
a__ : int = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
a__ : Optional[Any] = Image.open(requests.get(__a , stream=__a ).raw ).convert("RGB" )
return image
def UpperCamelCase_ ( __a ) -> Optional[Any]:
a__ : Any = []
# fmt: off
# vision encoder
rename_keys.append(("visual_encoder.cls_token", "vision_model.embeddings.class_embedding") )
rename_keys.append(("visual_encoder.pos_embed", "vision_model.embeddings.position_embedding") )
rename_keys.append(("visual_encoder.patch_embed.proj.weight", "vision_model.embeddings.patch_embedding.weight") )
rename_keys.append(("visual_encoder.patch_embed.proj.bias", "vision_model.embeddings.patch_embedding.bias") )
rename_keys.append(("ln_vision.weight", "vision_model.post_layernorm.weight") )
rename_keys.append(("ln_vision.bias", "vision_model.post_layernorm.bias") )
for i in range(config.vision_config.num_hidden_layers ):
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.weight''', f'''vision_model.encoder.layers.{i}.layer_norm1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.bias''', f'''vision_model.encoder.layers.{i}.layer_norm1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.weight''', f'''vision_model.encoder.layers.{i}.layer_norm2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.bias''', f'''vision_model.encoder.layers.{i}.layer_norm2.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.qkv.weight''', f'''vision_model.encoder.layers.{i}.self_attn.qkv.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.weight''', f'''vision_model.encoder.layers.{i}.self_attn.projection.weight''',) )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.bias''', f'''vision_model.encoder.layers.{i}.self_attn.projection.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc2.bias''') )
# QFormer
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.weight", "qformer.embeddings.layernorm.weight") )
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.bias", "qformer.embeddings.layernorm.bias") )
# fmt: on
return rename_keys
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Union[str, Any] = dct.pop(__a )
a__ : List[str] = val
def UpperCamelCase_ ( __a , __a ) -> Optional[Any]:
for i in range(config.vision_config.num_hidden_layers ):
# read in original q and v biases
a__ : Any = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.q_bias''' )
a__ : Tuple = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.v_bias''' )
# next, set bias in the state dict
a__ : str = torch.cat((q_bias, torch.zeros_like(__a , requires_grad=__a ), v_bias) )
a__ : int = qkv_bias
def UpperCamelCase_ ( __a ) -> Dict:
a__ : Tuple = 364 if "coco" in model_name else 224
a__ : int = InstructBlipVisionConfig(image_size=__a ).to_dict()
# make sure the models have proper bos_token_id and eos_token_id set (important for generation)
# seems like flan-T5 models don't have bos_token_id properly set?
if "t5-xl" in model_name:
a__ : Tuple = TaConfig.from_pretrained("google/flan-t5-xl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "t5-xxl" in model_name:
a__ : Dict = TaConfig.from_pretrained("google/flan-t5-xxl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "vicuna-7b" in model_name:
a__ : List[Any] = LlamaConfig.from_pretrained("decapoda-research/llama-7b-hf" , vocab_size=32_001 ).to_dict()
elif "vicuna-13b" in model_name:
a__ : Optional[int] = LlamaConfig.from_pretrained("decapoda-research/llama-13b-hf" , vocab_size=32_001 ).to_dict()
else:
raise ValueError("Model name not supported" )
# the authors add one special "[DEC]" token to the vocab of Q-Former, hence vocab size = 30522 + 1
a__ : Optional[Any] = InstructBlipQFormerConfig(vocab_size=30_523 ).to_dict()
a__ : Any = InstructBlipConfig(vision_config=__a , text_config=__a , qformer_config=__a )
return config, image_size
@torch.no_grad()
def UpperCamelCase_ ( __a , __a=None , __a=False ) -> int:
a__ : Tuple = AutoTokenizer.from_pretrained("bert-base-uncased" , truncation_side="left" )
qformer_tokenizer.add_special_tokens({"bos_token": "[DEC]"} )
if "t5" in model_name:
a__ : List[Any] = TaTokenizerFast.from_pretrained("google/flan-t5-xl" , truncation_side="left" )
elif "vicuna" in model_name:
# the following was used in the original implementation:
# tokenizer = LlamaTokenizer.from_pretrained("huggyllama/llama-7b", use_fast=False, truncation_side="left")
# tokenizer.add_special_tokens({"pad_token": "[PAD]"})
# tokenizer.add_special_tokens({"bos_token": "</s>"})
# tokenizer.add_special_tokens({"eos_token": "</s>"})
# tokenizer.add_special_tokens({"unk_token": "</s>"})
a__ : Union[str, Any] = LlamaTokenizerFast.from_pretrained(
"huggyllama/llama-7b" , truncation_side="left" , bos_token="</s>" , unk_token="</s>" )
tokenizer.add_special_tokens({"pad_token": "[PAD]"} )
a__, a__ : List[str] = get_blipa_config(__a )
a__ : Any = InstructBlipForConditionalGeneration(__a ).eval()
a__ : Dict = {
"instructblip-vicuna-7b": ("blip2_vicuna_instruct", "vicuna7b"),
"instructblip-vicuna-13b": ("blip2_vicuna_instruct", "vicuna13b"),
"instructblip-flan-t5-xl": ("blip2_t5_instruct", "flant5xl"),
"instructblip-flan-t5-xxl": ("blip2_t5_instruct", "flant5xxl"),
}
a__, a__ : Dict = model_name_to_original[model_name]
# load original model
print("Loading original model..." )
a__ : Optional[Any] = "cuda:1" if torch.cuda.is_available() else "cpu"
a__ : List[Any] = "cuda:2" if torch.cuda.is_available() else "cpu"
a__, a__, a__ : Tuple = load_model_and_preprocess(
name=__a , model_type=__a , is_eval=__a , device=__a )
original_model.eval()
print("Done!" )
# update state dict keys
a__ : Dict = original_model.state_dict()
a__ : Optional[int] = create_rename_keys(__a )
for src, dest in rename_keys:
rename_key(__a , __a , __a )
# some keys can be renamed efficiently
for key, val in state_dict.copy().items():
a__ : Optional[int] = state_dict.pop(__a )
if key.startswith("Qformer.bert" ):
a__ : List[Any] = key.replace("Qformer.bert" , "qformer" )
if "attention.self" in key:
a__ : Any = key.replace("self" , "attention" )
if "llm_proj" in key:
a__ : Dict = key.replace("llm_proj" , "language_projection" )
if "t5_proj" in key:
a__ : int = key.replace("t5_proj" , "language_projection" )
if key.startswith("llm_model" ):
a__ : List[str] = key.replace("llm_model" , "language_model" )
if key.startswith("t5" ):
a__ : str = key.replace("t5" , "language" )
a__ : Dict = val
# read in qv biases
read_in_q_v_bias(__a , __a )
# note: weights get loaded in torch.float32 by default
hf_model.load_state_dict(__a , strict=__a )
a__ : Union[str, Any] = load_demo_image()
a__ : int = "What is unusual about this image?"
# create processor
a__ : Any = BlipImageProcessor(
size={"height": image_size, "width": image_size} , image_mean=__a , image_std=__a )
a__ : Tuple = InstructBlipProcessor(
image_processor=__a , tokenizer=__a , qformer_tokenizer=__a , )
a__ : Tuple = processor(images=__a , text=__a , return_tensors="pt" ).to(__a )
# make sure processor creates exact same pixel values
a__ : Optional[int] = vis_processors["eval"](__a ).unsqueeze(0 ).to(__a )
a__ : Optional[Any] = inputs.pixel_values
assert torch.allclose(original_pixel_values.to(pixel_values.device ) , __a )
original_model.to(__a )
hf_model.to(__a )
with torch.no_grad():
if "vicuna" in model_name:
a__ : str = original_model({"image": original_pixel_values, "text_input": [prompt]} ).logits
a__ : List[str] = hf_model(**__a ).logits
else:
a__ : List[Any] = original_model(
{"image": original_pixel_values, "text_input": [prompt], "text_output": ["\n"]} ).logits
a__ : str = tokenizer("\n" , return_tensors="pt" ).input_ids.to(__a )
a__ : Dict = label_input_ids.masked_fill(label_input_ids == tokenizer.pad_token_id , -100 )
a__ : Any = hf_model(**__a , labels=__a ).logits
print("First values of original logits:" , original_logits[0, :3, :3] )
print("First values of HF logits:" , logits[0, :3, :3] )
# assert values
assert original_logits.shape == logits.shape
a__ : Tuple = 1e-4 if "vicuna" in model_name else 1e-5
assert torch.allclose(original_logits.to(logits.device ) , __a , atol=__a )
print("Looks ok!" )
print("Generating with original model..." )
a__ : Tuple = original_model.generate({"image": original_pixel_values, "prompt": prompt} , num_beams=5 )
# important: we need to cast the weights of the HF model to the appropriate type
print("Generating with HF model..." )
a__ : int = hf_model.generate(
**__a , do_sample=__a , num_beams=5 , max_length=256 , min_length=1 , top_p=0.9 , repetition_penalty=1.5 , length_penalty=1.0 , temperature=1 , )
if "vicuna" in model_name:
# convert output id 0 to 2 (eos_token_id)
# TODO add this in the generate method?
a__ : int = 2
print("Original generation:" , __a )
a__ : str = processor.batch_decode(__a , skip_special_tokens=__a )
a__ : str = [text.strip() for text in output_text]
print("HF generation:" , __a )
if pytorch_dump_folder_path is not None:
processor.save_pretrained(__a )
hf_model.save_pretrained(__a )
if push_to_hub:
processor.push_to_hub(f'''Salesforce/{model_name}''' )
hf_model.push_to_hub(f'''Salesforce/{model_name}''' )
if __name__ == "__main__":
UpperCamelCase : Any = argparse.ArgumentParser()
UpperCamelCase : Optional[int] = [
"""instructblip-vicuna-7b""",
"""instructblip-vicuna-13b""",
"""instructblip-flan-t5-xl""",
"""instructblip-flan-t5-xxl""",
]
parser.add_argument(
"""--model_name""",
default="""instructblip-flan-t5-xl""",
choices=choices,
type=str,
help="""Path to hf config.json of model to convert""",
)
parser.add_argument("""--pytorch_dump_folder_path""", default=None, type=str, help="""Path to the output PyTorch model.""")
parser.add_argument(
"""--push_to_hub""",
action="""store_true""",
help="""Whether to push the model and processor to the hub after converting""",
)
UpperCamelCase : Dict = parser.parse_args()
convert_blipa_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 37 | 1 |
from math import ceil
def UpperCamelCase_ ( __a = 1_001 ) -> int:
a__ : Optional[Any] = 1
for i in range(1 , int(ceil(n / 2.0 ) ) ):
a__ : List[str] = 2 * i + 1
a__ : Optional[int] = 2 * i
a__ : Dict = total + 4 * odd**2 - 6 * even
return total
if __name__ == "__main__":
import sys
if len(sys.argv) == 1:
print(solution())
else:
try:
UpperCamelCase : List[str] = int(sys.argv[1])
print(solution(n))
except ValueError:
print("""Invalid entry - please enter a number""")
| 37 |
def UpperCamelCase_ ( __a , __a ) -> Tuple:
a__ : Optional[int] = [0 for i in range(r + 1 )]
# nc0 = 1
a__ : Union[str, Any] = 1
for i in range(1 , n + 1 ):
# to compute current row from previous row.
a__ : Any = min(__a , __a )
while j > 0:
c[j] += c[j - 1]
j -= 1
return c[r]
print(binomial_coefficient(n=10, r=5))
| 37 | 1 |
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_vision_available
UpperCamelCase : List[str] = {
"""configuration_bridgetower""": [
"""BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP""",
"""BridgeTowerConfig""",
"""BridgeTowerTextConfig""",
"""BridgeTowerVisionConfig""",
],
"""processing_bridgetower""": ["""BridgeTowerProcessor"""],
}
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
UpperCamelCase : int = ["""BridgeTowerImageProcessor"""]
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
UpperCamelCase : str = [
"""BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST""",
"""BridgeTowerForContrastiveLearning""",
"""BridgeTowerForImageAndTextRetrieval""",
"""BridgeTowerForMaskedLM""",
"""BridgeTowerModel""",
"""BridgeTowerPreTrainedModel""",
]
if TYPE_CHECKING:
from .configuration_bridgetower import (
BRIDGETOWER_PRETRAINED_CONFIG_ARCHIVE_MAP,
BridgeTowerConfig,
BridgeTowerTextConfig,
BridgeTowerVisionConfig,
)
from .processing_bridgetower import BridgeTowerProcessor
try:
if not is_vision_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .image_processing_bridgetower import BridgeTowerImageProcessor
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_bridgetower import (
BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST,
BridgeTowerForContrastiveLearning,
BridgeTowerForImageAndTextRetrieval,
BridgeTowerForMaskedLM,
BridgeTowerModel,
BridgeTowerPreTrainedModel,
)
else:
import sys
UpperCamelCase : Optional[Any] = _LazyModule(__name__, globals()["""__file__"""], _import_structure)
| 37 |
import json
from typing import Dict, List, Optional, Tuple, Union
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import PaddingStrategy, logging
from .tokenization_led import LEDTokenizer
UpperCamelCase : Union[str, Any] = logging.get_logger(__name__)
UpperCamelCase : Optional[Any] = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Optional[Any] = {
"""vocab_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json""",
},
"""merges_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt""",
},
"""tokenizer_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json""",
},
}
UpperCamelCase : Dict = {
"""allenai/led-base-16384""": 1_6384,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = LEDTokenizer
_lowercase = ['input_ids', 'attention_mask']
def __init__( self : Tuple , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[str]=None , lowerCamelCase__ : Any=None , lowerCamelCase__ : int="replace" , lowerCamelCase__ : Union[str, Any]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Tuple="</s>" , lowerCamelCase__ : Optional[int]="<s>" , lowerCamelCase__ : str="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Any="<mask>" , lowerCamelCase__ : Optional[int]=False , lowerCamelCase__ : int=True , **lowerCamelCase__ : Union[str, Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : List[str] = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : Optional[Any] = add_prefix_space
a__ : List[str] = pre_tok_class(**lowerCamelCase__ )
a__ : Optional[int] = add_prefix_space
# the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__`
a__ : Any = "post_processor"
a__ : str = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Any = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Optional[Any] = tuple(state["sep"] )
if "cls" in state:
a__ : Optional[Any] = tuple(state["cls"] )
a__ : Optional[int] = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Dict = add_prefix_space
a__ : int = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : List[Any] = trim_offsets
a__ : List[str] = True
if changes_to_apply:
a__ : int = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : int = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
# Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Union[str, Any] ):
a__ : Any = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : Union[str, Any] = value
def _UpperCamelCase( self : Any , *lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : List[str] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
if is_split_into_words and not self.add_prefix_space:
raise ValueError(
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs." )
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Any , *lowerCamelCase__ : Dict , **lowerCamelCase__ : Optional[Any] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
if is_split_into_words and not self.add_prefix_space:
raise ValueError(
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs." )
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : List[str] = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[Any]=None ):
a__ : Any = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : List[str] = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Union[Dict[str, EncodedInput], BatchEncoding] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : PaddingStrategy = PaddingStrategy.DO_NOT_PAD , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[bool] = None , ):
a__ : str = super()._pad(
encoded_inputs=lowerCamelCase__ , max_length=lowerCamelCase__ , padding_strategy=lowerCamelCase__ , pad_to_multiple_of=lowerCamelCase__ , return_attention_mask=lowerCamelCase__ , )
# Load from model defaults
if return_attention_mask is None:
a__ : Optional[int] = "attention_mask" in self.model_input_names
if return_attention_mask and "global_attention_mask" in encoded_inputs:
a__ : Tuple = encoded_inputs[self.model_input_names[0]]
# `global_attention_mask` need to have the same length as other (sequential) inputs.
a__ : Dict = len(encoded_inputs["global_attention_mask"] ) != len(lowerCamelCase__ )
if needs_to_be_padded:
a__ : Union[str, Any] = len(lowerCamelCase__ ) - len(encoded_inputs["global_attention_mask"] )
if self.padding_side == "right":
# Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend`
a__ : List[Any] = (
encoded_inputs["global_attention_mask"] + [-1] * difference
)
elif self.padding_side == "left":
a__ : Any = [-1] * difference + encoded_inputs[
"global_attention_mask"
]
else:
raise ValueError("Invalid padding strategy:" + str(self.padding_side ) )
return encoded_inputs
| 37 | 1 |
import argparse
import logging
import os
import sys
import numpy as np
import onnxruntime
import torch
from bart_onnx.generation_onnx import BARTBeamSearchGenerator
from bart_onnx.reduce_onnx_size import remove_dup_initializers
import transformers
from transformers import BartForConditionalGeneration, BartTokenizer
logging.basicConfig(
format="""%(asctime)s | %(levelname)s | %(name)s | [%(filename)s:%(lineno)d] %(message)s""",
datefmt="""%Y-%m-%d %H:%M:%S""",
level=os.environ.get("""LOGLEVEL""", """INFO""").upper(),
stream=sys.stdout,
)
UpperCamelCase : List[str] = logging.getLogger(__name__)
UpperCamelCase : Optional[int] = {"""facebook/bart-base""": BartForConditionalGeneration}
UpperCamelCase : Optional[int] = {"""facebook/bart-base""": BartTokenizer}
def UpperCamelCase_ ( ) -> Optional[Any]:
a__ : Optional[Any] = argparse.ArgumentParser(description="Export Bart model + Beam Search to ONNX graph." )
parser.add_argument(
"--validation_file" , type=__a , default=__a , help="A csv or a json file containing the validation data." )
parser.add_argument(
"--max_length" , type=__a , default=5 , help="The maximum total input sequence length after tokenization." , )
parser.add_argument(
"--num_beams" , type=__a , default=__a , help=(
"Number of beams to use for evaluation. This argument will be "
"passed to ``model.generate``, which is used during ``evaluate`` and ``predict``."
) , )
parser.add_argument(
"--model_name_or_path" , type=__a , help="Path to pretrained model or model identifier from huggingface.co/models." , required=__a , )
parser.add_argument(
"--config_name" , type=__a , default=__a , help="Pretrained config name or path if not the same as model_name" , )
parser.add_argument(
"--device" , type=__a , default="cpu" , help="Device where the model will be run" , )
parser.add_argument("--output_file_path" , type=__a , default=__a , help="Where to store the final ONNX file." )
a__ : str = parser.parse_args()
return args
def UpperCamelCase_ ( __a , __a="cpu" ) -> List[Any]:
a__ : Any = model_dict[model_name].from_pretrained(__a ).to(__a )
a__ : List[str] = tokenizer_dict[model_name].from_pretrained(__a )
if model_name in ["facebook/bart-base"]:
a__ : Optional[int] = 0
a__ : Union[str, Any] = None
a__ : Union[str, Any] = 0
return huggingface_model, tokenizer
def UpperCamelCase_ ( __a , __a , __a , __a , __a ) -> Union[str, Any]:
model.eval()
a__ : int = None
a__ : str = torch.jit.script(BARTBeamSearchGenerator(__a ) )
with torch.no_grad():
a__ : int = "My friends are cool but they eat too many carbs."
a__ : str = tokenizer([ARTICLE_TO_SUMMARIZE] , max_length=1_024 , return_tensors="pt" ).to(model.device )
a__ : int = model.generate(
inputs["input_ids"] , attention_mask=inputs["attention_mask"] , num_beams=__a , max_length=__a , early_stopping=__a , decoder_start_token_id=model.config.decoder_start_token_id , )
torch.onnx.export(
__a , (
inputs["input_ids"],
inputs["attention_mask"],
num_beams,
max_length,
model.config.decoder_start_token_id,
) , __a , opset_version=14 , input_names=["input_ids", "attention_mask", "num_beams", "max_length", "decoder_start_token_id"] , output_names=["output_ids"] , dynamic_axes={
"input_ids": {0: "batch", 1: "seq"},
"output_ids": {0: "batch", 1: "seq_out"},
} , example_outputs=__a , )
logger.info("Model exported to {}".format(__a ) )
a__ : List[str] = remove_dup_initializers(os.path.abspath(__a ) )
logger.info("Deduplicated and optimized model written to {}".format(__a ) )
a__ : Dict = onnxruntime.InferenceSession(__a )
a__ : Tuple = ort_sess.run(
__a , {
"input_ids": inputs["input_ids"].cpu().numpy(),
"attention_mask": inputs["attention_mask"].cpu().numpy(),
"num_beams": np.array(__a ),
"max_length": np.array(__a ),
"decoder_start_token_id": np.array(model.config.decoder_start_token_id ),
} , )
np.testing.assert_allclose(summary_ids.cpu().numpy() , ort_out[0] , rtol=1e-3 , atol=1e-3 )
logger.info("Model outputs from torch and ONNX Runtime are similar." )
logger.info("Success." )
def UpperCamelCase_ ( ) -> Tuple:
a__ : int = parse_args()
a__ : str = 5
a__ : str = 4
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s" , datefmt="%m/%d/%Y %H:%M:%S" , level=logging.INFO , )
logger.setLevel(logging.INFO )
transformers.utils.logging.set_verbosity_error()
a__ : Optional[int] = torch.device(args.device )
a__, a__ : int = load_model_tokenizer(args.model_name_or_path , __a )
if model.config.decoder_start_token_id is None:
raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined" )
model.to(__a )
if args.max_length:
a__ : Optional[int] = args.max_length
if args.num_beams:
a__ : int = args.num_beams
if args.output_file_path:
a__ : str = args.output_file_path
else:
a__ : Optional[Any] = "BART.onnx"
logger.info("Exporting model to ONNX" )
export_and_validate_model(__a , __a , __a , __a , __a )
if __name__ == "__main__":
main()
| 37 |
import json
from typing import List, Optional, Tuple
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import logging
from .tokenization_roberta import RobertaTokenizer
UpperCamelCase : Any = logging.get_logger(__name__)
UpperCamelCase : Any = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Union[str, Any] = {
"""vocab_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/vocab.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/vocab.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/vocab.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/vocab.json""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/vocab.json""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/vocab.json"""
),
},
"""merges_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/merges.txt""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/merges.txt""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/merges.txt""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/merges.txt""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/merges.txt""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/merges.txt"""
),
},
"""tokenizer_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/tokenizer.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/tokenizer.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/tokenizer.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/tokenizer.json""",
"""roberta-base-openai-detector""": (
"""https://huggingface.co/roberta-base-openai-detector/resolve/main/tokenizer.json"""
),
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/tokenizer.json"""
),
},
}
UpperCamelCase : List[str] = {
"""roberta-base""": 512,
"""roberta-large""": 512,
"""roberta-large-mnli""": 512,
"""distilroberta-base""": 512,
"""roberta-base-openai-detector""": 512,
"""roberta-large-openai-detector""": 512,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = ['input_ids', 'attention_mask']
_lowercase = RobertaTokenizer
def __init__( self : List[str] , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[Any]=None , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]="replace" , lowerCamelCase__ : List[str]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Any="</s>" , lowerCamelCase__ : Any="<s>" , lowerCamelCase__ : int="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Tuple="<mask>" , lowerCamelCase__ : Any=False , lowerCamelCase__ : Dict=True , **lowerCamelCase__ : Optional[Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Any = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : int = add_prefix_space
a__ : Tuple = pre_tok_class(**lowerCamelCase__ )
a__ : str = add_prefix_space
a__ : Tuple = "post_processor"
a__ : Dict = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Tuple = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Tuple = tuple(state["sep"] )
if "cls" in state:
a__ : str = tuple(state["cls"] )
a__ : str = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : str = add_prefix_space
a__ : Any = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : int = trim_offsets
a__ : Dict = True
if changes_to_apply:
a__ : Union[str, Any] = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : str = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Tuple ):
a__ : List[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : List[str] = value
def _UpperCamelCase( self : Union[str, Any] , *lowerCamelCase__ : int , **lowerCamelCase__ : int ):
a__ : Optional[int] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Tuple , *lowerCamelCase__ : Dict , **lowerCamelCase__ : List[str] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : str , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : int = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[int]=None ):
a__ : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : Tuple = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
| 37 | 1 |
import argparse
import shutil
import time
from json import JSONDecodeError
from logging import getLogger
from pathlib import Path
from typing import Dict, List
import torch
from torch.utils.data import DataLoader
from tqdm import tqdm
from transformers import AutoModelForSeqaSeqLM, AutoTokenizer
from utils import (
SeqaSeqDataset,
calculate_bleu,
calculate_rouge,
chunks,
lmap,
load_json,
parse_numeric_n_bool_cl_kwargs,
save_json,
use_task_specific_params,
write_txt_file,
)
UpperCamelCase : Union[str, Any] = getLogger(__name__)
def UpperCamelCase_ ( __a , __a , __a , __a = 8 , __a = 1_024 , __a="val" , __a=None , __a=False , __a="summarization" , __a=None , __a=1 , __a = None , __a="" , **__a , ) -> Dict:
a__ : List[str] = str(__a )
assert local_rank is not None
torch.distributed.init_process_group(backend="nccl" , rank=__a )
a__ : Tuple = Path(__a )
a__ : Tuple = save_dir.joinpath(f'''rank_{local_rank}_output.json''' )
torch.cuda.set_device(__a )
a__ : Dict = AutoModelForSeqaSeqLM.from_pretrained(__a ).cuda()
if fpaa:
a__ : List[str] = model.half()
# determine if we need to increase num_beams
use_task_specific_params(__a , __a ) # update config with task specific params
a__ : Optional[int] = generate_kwargs.pop("num_beams" , model.config.num_beams ) # AttributeError risk?
if num_return_sequences > num_beams:
a__ : Tuple = num_return_sequences
a__ : int = AutoTokenizer.from_pretrained(__a )
logger.info(f'''Inferred tokenizer type: {tokenizer.__class__}''' ) # if this is wrong, check config.model_type.
if max_source_length is None:
a__ : Optional[int] = tokenizer.model_max_length
if prefix is None:
a__ : List[str] = prefix or getattr(model.config , "prefix" , "" ) or ""
a__ : Tuple = SeqaSeqDataset(
__a , __a , __a , max_target_length=1_024 , type_path=__a , n_obs=__a , prefix=__a , **__a , )
# I set shuffle=True for a more accurate progress bar.
# If all the longest samples are first, the prog bar estimate is too high at the beginning.
a__ : Any = ds.make_sortish_sampler(__a , distributed=__a , add_extra_examples=__a , shuffle=__a )
a__ : List[str] = DataLoader(__a , sampler=__a , batch_size=__a , collate_fn=ds.collate_fn )
a__ : List[str] = []
for batch in tqdm(__a ):
a__ : List[str] = model.generate(
input_ids=batch["input_ids"].to(model.device ) , attention_mask=batch["attention_mask"].to(model.device ) , num_return_sequences=__a , num_beams=__a , **__a , )
a__ : int = tokenizer.batch_decode(__a , skip_special_tokens=__a , clean_up_tokenization_spaces=__a )
a__ : Optional[int] = batch["ids"]
if num_return_sequences > 1:
a__ : Any = chunks(__a , __a ) # batch size chunks, each of size num_return_seq
for i, pred in enumerate(__a ):
results.append({"pred": pred, "id": ids[i].item()} )
save_json(__a , __a )
return results, sampler.num_replicas
def UpperCamelCase_ ( ) -> List[str]:
a__ : Any = argparse.ArgumentParser(
epilog="Unspecified args like --num_beams=2 --decoder_start_token_id=4 are passed to model.generate" )
parser.add_argument("--data_dir" , type=__a , help="like cnn_dm/test.source" )
parser.add_argument(
"--model_name" , type=__a , help="like facebook/bart-large-cnn,t5-base, etc." , default="sshleifer/distilbart-xsum-12-3" , )
parser.add_argument("--save_dir" , type=__a , help="where to save" , default="tmp_gen" )
parser.add_argument("--max_source_length" , type=__a , default=__a )
parser.add_argument(
"--type_path" , type=__a , default="test" , help="which subset to evaluate typically train/val/test" )
parser.add_argument("--task" , type=__a , default="summarization" , help="used for task_specific_params + metrics" )
parser.add_argument("--bs" , type=__a , default=8 , required=__a , help="batch size" )
parser.add_argument(
"--local_rank" , type=__a , default=-1 , required=__a , help="should be passed by distributed.launch" )
parser.add_argument(
"--n_obs" , type=__a , default=__a , required=__a , help="How many observations. Defaults to all." )
parser.add_argument(
"--num_return_sequences" , type=__a , default=1 , required=__a , help="How many sequences to return" )
parser.add_argument(
"--sync_timeout" , type=__a , default=600 , required=__a , help="How long should master process wait for other processes to finish." , )
parser.add_argument("--src_lang" , type=__a , default=__a , required=__a )
parser.add_argument("--tgt_lang" , type=__a , default=__a , required=__a )
parser.add_argument(
"--prefix" , type=__a , required=__a , default=__a , help="will be added to the begininng of src examples" )
parser.add_argument("--fp16" , action="store_true" )
parser.add_argument("--debug" , action="store_true" )
a__ : Tuple = time.time()
a__, a__ : str = parser.parse_known_args()
a__ : Any = parse_numeric_n_bool_cl_kwargs(__a )
if generate_kwargs and args.local_rank <= 0:
print(f'''parsed the following generate kwargs: {generate_kwargs}''' )
a__ : int = Path(args.save_dir + "_tmp" )
Path(__a ).mkdir(exist_ok=__a ) # this handles locking.
a__ : Any = list(json_save_dir.glob("rank_*.json" ) )
if intermediate_files:
raise ValueError(f'''Found files at {json_save_dir} please move or remove them.''' )
# In theory, a node could finish and save before another node hits this. If this happens, we can address later.
a__ : Dict = {}
if args.src_lang is not None:
a__ : Any = args.src_lang
if args.tgt_lang is not None:
a__ : Tuple = args.tgt_lang
Path(args.save_dir ).mkdir(exist_ok=__a )
a__, a__ : int = eval_data_dir(
args.data_dir , __a , args.model_name , type_path=args.type_path , bs=args.bs , fpaa=args.fpaa , task=args.task , local_rank=args.local_rank , n_obs=args.n_obs , max_source_length=args.max_source_length , num_return_sequences=args.num_return_sequences , prefix=args.prefix , dataset_kwargs=__a , **__a , )
if args.local_rank <= 0:
a__ : Tuple = Path(args.save_dir )
save_dir.mkdir(exist_ok=__a )
a__ : Optional[int] = gather_results_from_each_node(__a , __a , args.sync_timeout )
a__ : Tuple = combine_partial_results(__a )
if args.num_return_sequences > 1:
a__ : Dict = save_dir.joinpath("pseudolabel_results.json" )
print(f'''Saving aggregated results at {save_path}, intermediate in {json_save_dir}/''' )
save_json(__a , __a )
return
a__ : str = Path(args.data_dir ).joinpath(args.type_path + ".target" )
with open(__a ) as f:
a__ : Union[str, Any] = [x.rstrip() for x in f.readlines()][: len(__a )]
# Calculate metrics, save metrics, and save _generations.txt
a__ : Dict = "translation" in args.task
a__ : str = calculate_bleu if calc_bleu else calculate_rouge
a__ : str = "bleu" if calc_bleu else "rouge"
a__ : Dict = score_fn(__a , __a )
a__ : List[str] = len(__a )
a__ : Any = time.time() - start_time
a__ : Optional[int] = round(runtime / metrics["n_obs"] , 4 )
a__ : int = num_replicas
# TODO(@stas00): add whatever metadata to metrics
a__ : Dict = save_dir.joinpath(f'''{args.type_path}_{metric_name}.json''' )
save_json(__a , __a , indent=__a )
print(__a )
write_txt_file(__a , save_dir.joinpath(f'''{args.type_path}_generations.txt''' ) )
if args.debug:
write_txt_file(__a , save_dir.joinpath(f'''{args.type_path}.target''' ) )
else:
shutil.rmtree(__a )
def UpperCamelCase_ ( __a ) -> List:
a__ : List[str] = []
for partial_result in partial_results:
records.extend(__a )
a__ : List[str] = sorted(__a , key=lambda __a : x["id"] )
a__ : Union[str, Any] = [x["pred"] for x in records]
return preds
def UpperCamelCase_ ( __a , __a , __a ) -> List[Dict[str, List]]:
# WAIT FOR lots of .json files
a__ : List[Any] = time.time()
logger.info("waiting for all nodes to finish" )
a__ : Dict = None
while (time.time() - start_wait) < timeout:
a__ : Optional[Any] = list(save_dir.glob("rank_*.json" ) )
if len(__a ) < num_replicas:
continue
try:
# make sure all json files are fully saved
a__ : Union[str, Any] = lmap(__a , __a )
return json_data
except JSONDecodeError:
continue
else:
raise TimeoutError("Rank 0 gave up on waiting for other processes" )
# Unreachable
if __name__ == "__main__":
# Usage for MT:
run_generate()
| 37 |
from statistics import mean, stdev
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : List[str] = min(__a )
a__ : str = max(__a )
# normalize data
return [round((x - x_min) / (x_max - x_min) , __a ) for x in data]
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : str = mean(__a )
a__ : List[str] = stdev(__a )
# standardize data
return [round((x - mu) / (sigma) , __a ) for x in data]
| 37 | 1 |
import collections
import os
from typing import List, Optional, Tuple
from transformers.utils import is_jieba_available, requires_backends
if is_jieba_available():
import jieba
from ...tokenization_utils import PreTrainedTokenizer
from ...utils import logging
UpperCamelCase : Union[str, Any] = logging.get_logger(__name__)
UpperCamelCase : Tuple = {"""vocab_file""": """vocab.txt"""}
UpperCamelCase : Union[str, Any] = {
"""vocab_file""": {
"""openbmb/cpm-ant-10b""": """https://huggingface.co/openbmb/cpm-ant-10b/blob/main/vocab.txt""",
},
}
UpperCamelCase : int = {
"""openbmb/cpm-ant-10b""": 1024,
}
def UpperCamelCase_ ( __a ) -> Tuple:
a__ : List[str] = collections.OrderedDict()
with open(__a , "r" , encoding="utf-8" ) as reader:
a__ : Optional[int] = reader.readlines()
for index, token in enumerate(__a ):
a__ : List[str] = token.rstrip("\n" )
a__ : Tuple = index
return vocab
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : int , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Union[str, Any]="<unk>" , lowerCamelCase__ : Any=200 ):
a__ : Tuple = vocab
a__ : Tuple = unk_token
a__ : List[Any] = max_input_chars_per_word
def _UpperCamelCase( self : int , lowerCamelCase__ : List[str] ):
a__ : List[str] = list(lowerCamelCase__ )
if len(lowerCamelCase__ ) > self.max_input_chars_per_word:
return [self.unk_token]
a__ : Optional[int] = 0
a__ : Tuple = []
while start < len(lowerCamelCase__ ):
a__ : Union[str, Any] = len(lowerCamelCase__ )
a__ : List[str] = None
while start < end:
a__ : int = "".join(chars[start:end] )
if substr in self.vocab:
a__ : Optional[Any] = substr
break
end -= 1
if cur_substr is None:
sub_tokens.append(self.unk_token )
start += 1
else:
sub_tokens.append(lowerCamelCase__ )
a__ : List[Any] = end
return sub_tokens
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = ['input_ids', 'attention_mask']
_lowercase = False
def __init__( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[int]="<d>" , lowerCamelCase__ : List[Any]="</d>" , lowerCamelCase__ : Union[str, Any]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : List[Any]="<pad>" , lowerCamelCase__ : str="<unk>" , lowerCamelCase__ : int="</n>" , lowerCamelCase__ : Optional[Any]="</_>" , lowerCamelCase__ : Any="left" , **lowerCamelCase__ : Optional[Any] , ):
requires_backends(self , ["jieba"] )
super().__init__(
bod_token=lowerCamelCase__ , eod_token=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , line_token=lowerCamelCase__ , space_token=lowerCamelCase__ , padding_side=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : Union[str, Any] = bod_token
a__ : Optional[Any] = eod_token
a__ : Any = load_vocab(lowerCamelCase__ )
a__ : Dict = self.encoder[space_token]
a__ : List[str] = self.encoder[line_token]
del self.encoder[space_token]
del self.encoder[line_token]
a__ : Optional[Any] = collections.OrderedDict(sorted(self.encoder.items() , key=lambda lowerCamelCase__ : x[1] ) )
a__ : Union[str, Any] = {v: k for k, v in self.encoder.items()}
a__ : Optional[int] = WordpieceTokenizer(vocab=self.encoder , unk_token=self.unk_token )
@property
def _UpperCamelCase( self : List[str] ):
return self.encoder[self.bod_token]
@property
def _UpperCamelCase( self : Union[str, Any] ):
return self.encoder[self.eod_token]
@property
def _UpperCamelCase( self : Tuple ):
return self.encoder["\n"]
@property
def _UpperCamelCase( self : Optional[int] ):
return len(self.encoder )
def _UpperCamelCase( self : Dict ):
return dict(self.encoder , **self.added_tokens_encoder )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Optional[Any] ):
a__ : Optional[Any] = []
for x in jieba.cut(lowerCamelCase__ , cut_all=lowerCamelCase__ ):
output_tokens.extend(self.wordpiece_tokenizer.tokenize(lowerCamelCase__ ) )
return output_tokens
def _UpperCamelCase( self : str , lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Tuple ):
a__ : Tuple = [i for i in token_ids if i >= 0]
a__ : List[str] = [
x for x in token_ids if x != self.pad_token_id and x != self.eos_token_id and x != self.bos_token_id
]
return super()._decode(lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Optional[int] ):
return token in self.encoder
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : List[str] ):
return "".join(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : str ):
return self.encoder.get(lowerCamelCase__ , self.encoder.get(self.unk_token ) )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Tuple ):
return self.decoder.get(lowerCamelCase__ , self.unk_token )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
if os.path.isdir(lowerCamelCase__ ):
a__ : Optional[int] = os.path.join(
lowerCamelCase__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] )
else:
a__ : Optional[int] = (filename_prefix + "-" if filename_prefix else "") + save_directory
a__ : Dict = 0
if " " in self.encoder:
a__ : int = self.encoder[" "]
del self.encoder[" "]
if "\n" in self.encoder:
a__ : int = self.encoder["\n"]
del self.encoder["\n"]
a__ : Optional[int] = collections.OrderedDict(sorted(self.encoder.items() , key=lambda lowerCamelCase__ : x[1] ) )
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as writer:
for token, token_index in self.encoder.items():
if index != token_index:
logger.warning(
f'''Saving vocabulary to {vocab_file}: vocabulary indices are not consecutive.'''
" Please check that the vocabulary is not corrupted!" )
a__ : List[Any] = token_index
writer.write(token + "\n" )
index += 1
return (vocab_file,)
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : List[int] , lowerCamelCase__ : List[int] = None ):
if token_ids_a is None:
return [self.bos_token_id] + token_ids_a
return [self.bos_token_id] + token_ids_a + [self.bos_token_id] + token_ids_a
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None , lowerCamelCase__ : bool = False ):
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=lowerCamelCase__ , token_ids_a=lowerCamelCase__ , already_has_special_tokens=lowerCamelCase__ )
if token_ids_a is not None:
return [1] + ([0] * len(lowerCamelCase__ )) + [1] + ([0] * len(lowerCamelCase__ ))
return [1] + ([0] * len(lowerCamelCase__ ))
| 37 |
def UpperCamelCase_ ( __a = 50 ) -> int:
a__ : Tuple = [[0] * 3 for _ in range(length + 1 )]
for row_length in range(length + 1 ):
for tile_length in range(2 , 5 ):
for tile_start in range(row_length - tile_length + 1 ):
different_colour_ways_number[row_length][tile_length - 2] += (
different_colour_ways_number[row_length - tile_start - tile_length][
tile_length - 2
]
+ 1
)
return sum(different_colour_ways_number[length] )
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 | 1 |
import argparse
import requests
import torch
# pip3 install salesforce-lavis
# I'm actually installing a slightly modified version: pip3 install git+https://github.com/nielsrogge/LAVIS.git@fix_lavis_float32 (there's also the fix_lavis branch)
# also note: to convert Vicuna checkpoints, we had to include /home/niels/python_projects/checkpoints/FastChat/vicuna-7b in lavis/configs/models/blip2/blip2_instruct_vicuna7b.yaml
# same for Vicuna-13b
from lavis.models import load_model_and_preprocess
from PIL import Image
from transformers import (
AutoTokenizer,
BlipImageProcessor,
InstructBlipConfig,
InstructBlipForConditionalGeneration,
InstructBlipProcessor,
InstructBlipQFormerConfig,
InstructBlipVisionConfig,
LlamaConfig,
LlamaTokenizerFast,
TaConfig,
TaTokenizerFast,
)
from transformers.utils.constants import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD
def UpperCamelCase_ ( ) -> int:
a__ : int = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
a__ : Optional[Any] = Image.open(requests.get(__a , stream=__a ).raw ).convert("RGB" )
return image
def UpperCamelCase_ ( __a ) -> Optional[Any]:
a__ : Any = []
# fmt: off
# vision encoder
rename_keys.append(("visual_encoder.cls_token", "vision_model.embeddings.class_embedding") )
rename_keys.append(("visual_encoder.pos_embed", "vision_model.embeddings.position_embedding") )
rename_keys.append(("visual_encoder.patch_embed.proj.weight", "vision_model.embeddings.patch_embedding.weight") )
rename_keys.append(("visual_encoder.patch_embed.proj.bias", "vision_model.embeddings.patch_embedding.bias") )
rename_keys.append(("ln_vision.weight", "vision_model.post_layernorm.weight") )
rename_keys.append(("ln_vision.bias", "vision_model.post_layernorm.bias") )
for i in range(config.vision_config.num_hidden_layers ):
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.weight''', f'''vision_model.encoder.layers.{i}.layer_norm1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.bias''', f'''vision_model.encoder.layers.{i}.layer_norm1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.weight''', f'''vision_model.encoder.layers.{i}.layer_norm2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.bias''', f'''vision_model.encoder.layers.{i}.layer_norm2.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.qkv.weight''', f'''vision_model.encoder.layers.{i}.self_attn.qkv.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.weight''', f'''vision_model.encoder.layers.{i}.self_attn.projection.weight''',) )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.bias''', f'''vision_model.encoder.layers.{i}.self_attn.projection.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc2.bias''') )
# QFormer
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.weight", "qformer.embeddings.layernorm.weight") )
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.bias", "qformer.embeddings.layernorm.bias") )
# fmt: on
return rename_keys
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Union[str, Any] = dct.pop(__a )
a__ : List[str] = val
def UpperCamelCase_ ( __a , __a ) -> Optional[Any]:
for i in range(config.vision_config.num_hidden_layers ):
# read in original q and v biases
a__ : Any = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.q_bias''' )
a__ : Tuple = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.v_bias''' )
# next, set bias in the state dict
a__ : str = torch.cat((q_bias, torch.zeros_like(__a , requires_grad=__a ), v_bias) )
a__ : int = qkv_bias
def UpperCamelCase_ ( __a ) -> Dict:
a__ : Tuple = 364 if "coco" in model_name else 224
a__ : int = InstructBlipVisionConfig(image_size=__a ).to_dict()
# make sure the models have proper bos_token_id and eos_token_id set (important for generation)
# seems like flan-T5 models don't have bos_token_id properly set?
if "t5-xl" in model_name:
a__ : Tuple = TaConfig.from_pretrained("google/flan-t5-xl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "t5-xxl" in model_name:
a__ : Dict = TaConfig.from_pretrained("google/flan-t5-xxl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "vicuna-7b" in model_name:
a__ : List[Any] = LlamaConfig.from_pretrained("decapoda-research/llama-7b-hf" , vocab_size=32_001 ).to_dict()
elif "vicuna-13b" in model_name:
a__ : Optional[int] = LlamaConfig.from_pretrained("decapoda-research/llama-13b-hf" , vocab_size=32_001 ).to_dict()
else:
raise ValueError("Model name not supported" )
# the authors add one special "[DEC]" token to the vocab of Q-Former, hence vocab size = 30522 + 1
a__ : Optional[Any] = InstructBlipQFormerConfig(vocab_size=30_523 ).to_dict()
a__ : Any = InstructBlipConfig(vision_config=__a , text_config=__a , qformer_config=__a )
return config, image_size
@torch.no_grad()
def UpperCamelCase_ ( __a , __a=None , __a=False ) -> int:
a__ : Tuple = AutoTokenizer.from_pretrained("bert-base-uncased" , truncation_side="left" )
qformer_tokenizer.add_special_tokens({"bos_token": "[DEC]"} )
if "t5" in model_name:
a__ : List[Any] = TaTokenizerFast.from_pretrained("google/flan-t5-xl" , truncation_side="left" )
elif "vicuna" in model_name:
# the following was used in the original implementation:
# tokenizer = LlamaTokenizer.from_pretrained("huggyllama/llama-7b", use_fast=False, truncation_side="left")
# tokenizer.add_special_tokens({"pad_token": "[PAD]"})
# tokenizer.add_special_tokens({"bos_token": "</s>"})
# tokenizer.add_special_tokens({"eos_token": "</s>"})
# tokenizer.add_special_tokens({"unk_token": "</s>"})
a__ : Union[str, Any] = LlamaTokenizerFast.from_pretrained(
"huggyllama/llama-7b" , truncation_side="left" , bos_token="</s>" , unk_token="</s>" )
tokenizer.add_special_tokens({"pad_token": "[PAD]"} )
a__, a__ : List[str] = get_blipa_config(__a )
a__ : Any = InstructBlipForConditionalGeneration(__a ).eval()
a__ : Dict = {
"instructblip-vicuna-7b": ("blip2_vicuna_instruct", "vicuna7b"),
"instructblip-vicuna-13b": ("blip2_vicuna_instruct", "vicuna13b"),
"instructblip-flan-t5-xl": ("blip2_t5_instruct", "flant5xl"),
"instructblip-flan-t5-xxl": ("blip2_t5_instruct", "flant5xxl"),
}
a__, a__ : Dict = model_name_to_original[model_name]
# load original model
print("Loading original model..." )
a__ : Optional[Any] = "cuda:1" if torch.cuda.is_available() else "cpu"
a__ : List[Any] = "cuda:2" if torch.cuda.is_available() else "cpu"
a__, a__, a__ : Tuple = load_model_and_preprocess(
name=__a , model_type=__a , is_eval=__a , device=__a )
original_model.eval()
print("Done!" )
# update state dict keys
a__ : Dict = original_model.state_dict()
a__ : Optional[int] = create_rename_keys(__a )
for src, dest in rename_keys:
rename_key(__a , __a , __a )
# some keys can be renamed efficiently
for key, val in state_dict.copy().items():
a__ : Optional[int] = state_dict.pop(__a )
if key.startswith("Qformer.bert" ):
a__ : List[Any] = key.replace("Qformer.bert" , "qformer" )
if "attention.self" in key:
a__ : Any = key.replace("self" , "attention" )
if "llm_proj" in key:
a__ : Dict = key.replace("llm_proj" , "language_projection" )
if "t5_proj" in key:
a__ : int = key.replace("t5_proj" , "language_projection" )
if key.startswith("llm_model" ):
a__ : List[str] = key.replace("llm_model" , "language_model" )
if key.startswith("t5" ):
a__ : str = key.replace("t5" , "language" )
a__ : Dict = val
# read in qv biases
read_in_q_v_bias(__a , __a )
# note: weights get loaded in torch.float32 by default
hf_model.load_state_dict(__a , strict=__a )
a__ : Union[str, Any] = load_demo_image()
a__ : int = "What is unusual about this image?"
# create processor
a__ : Any = BlipImageProcessor(
size={"height": image_size, "width": image_size} , image_mean=__a , image_std=__a )
a__ : Tuple = InstructBlipProcessor(
image_processor=__a , tokenizer=__a , qformer_tokenizer=__a , )
a__ : Tuple = processor(images=__a , text=__a , return_tensors="pt" ).to(__a )
# make sure processor creates exact same pixel values
a__ : Optional[int] = vis_processors["eval"](__a ).unsqueeze(0 ).to(__a )
a__ : Optional[Any] = inputs.pixel_values
assert torch.allclose(original_pixel_values.to(pixel_values.device ) , __a )
original_model.to(__a )
hf_model.to(__a )
with torch.no_grad():
if "vicuna" in model_name:
a__ : str = original_model({"image": original_pixel_values, "text_input": [prompt]} ).logits
a__ : List[str] = hf_model(**__a ).logits
else:
a__ : List[Any] = original_model(
{"image": original_pixel_values, "text_input": [prompt], "text_output": ["\n"]} ).logits
a__ : str = tokenizer("\n" , return_tensors="pt" ).input_ids.to(__a )
a__ : Dict = label_input_ids.masked_fill(label_input_ids == tokenizer.pad_token_id , -100 )
a__ : Any = hf_model(**__a , labels=__a ).logits
print("First values of original logits:" , original_logits[0, :3, :3] )
print("First values of HF logits:" , logits[0, :3, :3] )
# assert values
assert original_logits.shape == logits.shape
a__ : Tuple = 1e-4 if "vicuna" in model_name else 1e-5
assert torch.allclose(original_logits.to(logits.device ) , __a , atol=__a )
print("Looks ok!" )
print("Generating with original model..." )
a__ : Tuple = original_model.generate({"image": original_pixel_values, "prompt": prompt} , num_beams=5 )
# important: we need to cast the weights of the HF model to the appropriate type
print("Generating with HF model..." )
a__ : int = hf_model.generate(
**__a , do_sample=__a , num_beams=5 , max_length=256 , min_length=1 , top_p=0.9 , repetition_penalty=1.5 , length_penalty=1.0 , temperature=1 , )
if "vicuna" in model_name:
# convert output id 0 to 2 (eos_token_id)
# TODO add this in the generate method?
a__ : int = 2
print("Original generation:" , __a )
a__ : str = processor.batch_decode(__a , skip_special_tokens=__a )
a__ : str = [text.strip() for text in output_text]
print("HF generation:" , __a )
if pytorch_dump_folder_path is not None:
processor.save_pretrained(__a )
hf_model.save_pretrained(__a )
if push_to_hub:
processor.push_to_hub(f'''Salesforce/{model_name}''' )
hf_model.push_to_hub(f'''Salesforce/{model_name}''' )
if __name__ == "__main__":
UpperCamelCase : Any = argparse.ArgumentParser()
UpperCamelCase : Optional[int] = [
"""instructblip-vicuna-7b""",
"""instructblip-vicuna-13b""",
"""instructblip-flan-t5-xl""",
"""instructblip-flan-t5-xxl""",
]
parser.add_argument(
"""--model_name""",
default="""instructblip-flan-t5-xl""",
choices=choices,
type=str,
help="""Path to hf config.json of model to convert""",
)
parser.add_argument("""--pytorch_dump_folder_path""", default=None, type=str, help="""Path to the output PyTorch model.""")
parser.add_argument(
"""--push_to_hub""",
action="""store_true""",
help="""Whether to push the model and processor to the hub after converting""",
)
UpperCamelCase : Dict = parser.parse_args()
convert_blipa_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 37 |
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[str] ):
a__ : str = name
a__ : Optional[int] = value
a__ : Dict = weight
def __repr__( self : Union[str, Any] ):
return f'''{self.__class__.__name__}({self.name}, {self.value}, {self.weight})'''
def _UpperCamelCase( self : Dict ):
return self.value
def _UpperCamelCase( self : Optional[Any] ):
return self.name
def _UpperCamelCase( self : Optional[Any] ):
return self.weight
def _UpperCamelCase( self : Optional[int] ):
return self.value / self.weight
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = []
for i in range(len(__a ) ):
menu.append(Things(name[i] , value[i] , weight[i] ) )
return menu
def UpperCamelCase_ ( __a , __a , __a ) -> Union[str, Any]:
a__ : List[str] = sorted(__a , key=__a , reverse=__a )
a__ : List[Any] = []
a__, a__ : Union[str, Any] = 0.0, 0.0
for i in range(len(__a ) ):
if (total_cost + items_copy[i].get_weight()) <= max_cost:
result.append(items_copy[i] )
total_cost += items_copy[i].get_weight()
total_value += items_copy[i].get_value()
return (result, total_value)
def UpperCamelCase_ ( ) -> Union[str, Any]:
pass
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 | 1 |
def UpperCamelCase_ ( __a , __a ) -> list:
a__ : str = word.split()
def justify(__a , __a , __a ) -> str:
a__ : List[str] = max_width - width
a__ : str = len(__a )
if len(__a ) == 1:
# if there is only word in line
# just insert overall_spaces_count for the remainder of line
return line[0] + " " * overall_spaces_count
else:
a__ : Dict = words_count - 1
# num_spaces_between_words_list[i] : tells you to insert
# num_spaces_between_words_list[i] spaces
# after word on line[i]
a__ : str = spaces_to_insert_between_words * [
overall_spaces_count // spaces_to_insert_between_words
]
a__ : List[Any] = (
overall_spaces_count % spaces_to_insert_between_words
)
# distribute spaces via round robin to the left words
for i in range(__a ):
num_spaces_between_words_list[i] += 1
a__ : Union[str, Any] = []
for i in range(__a ):
# add the word
aligned_words_list.append(line[i] )
# add the spaces to insert
aligned_words_list.append(num_spaces_between_words_list[i] * " " )
# just add the last word to the sentence
aligned_words_list.append(line[-1] )
# join the aligned words list to form a justified line
return "".join(__a )
a__ : Any = []
a__ : list[str] = []
a__ : Any = 0
for word in words:
if width + len(__a ) + len(__a ) <= max_width:
# keep adding words until we can fill out max_width
# width = sum of length of all words (without overall_spaces_count)
# len(word) = length of current word
# len(line) = number of overall_spaces_count to insert between words
line.append(__a )
width += len(__a )
else:
# justify the line and add it to result
answer.append(justify(__a , __a , __a ) )
# reset new line and new width
a__, a__ : Optional[int] = [word], len(__a )
a__ : int = max_width - width - len(__a )
answer.append(" ".join(__a ) + (remaining_spaces + 1) * " " )
return answer
if __name__ == "__main__":
from doctest import testmod
testmod()
| 37 |
import multiprocessing
from typing import TYPE_CHECKING, Optional, Union
from .. import Dataset, Features, config
from ..formatting import query_table
from ..packaged_modules.sql.sql import Sql
from ..utils import logging
from .abc import AbstractDatasetInputStream
if TYPE_CHECKING:
import sqlitea
import sqlalchemy
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Dict , lowerCamelCase__ : Union[str, "sqlalchemy.sql.Selectable"] , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[Features] = None , lowerCamelCase__ : str = None , lowerCamelCase__ : bool = False , **lowerCamelCase__ : Optional[int] , ):
super().__init__(features=lowerCamelCase__ , cache_dir=lowerCamelCase__ , keep_in_memory=lowerCamelCase__ , **lowerCamelCase__ )
a__ : str = Sql(
cache_dir=lowerCamelCase__ , features=lowerCamelCase__ , sql=lowerCamelCase__ , con=lowerCamelCase__ , **lowerCamelCase__ , )
def _UpperCamelCase( self : Tuple ):
a__ : Optional[Any] = None
a__ : Dict = None
a__ : Union[str, Any] = None
a__ : Union[str, Any] = None
self.builder.download_and_prepare(
download_config=lowerCamelCase__ , download_mode=lowerCamelCase__ , verification_mode=lowerCamelCase__ , base_path=lowerCamelCase__ , )
# Build dataset for splits
a__ : List[str] = self.builder.as_dataset(
split="train" , verification_mode=lowerCamelCase__ , in_memory=self.keep_in_memory )
return dataset
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : Dataset , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[int] = None , **lowerCamelCase__ : Optional[Any] , ):
if num_proc is not None and num_proc <= 0:
raise ValueError(f'''num_proc {num_proc} must be an integer > 0.''' )
a__ : Any = dataset
a__ : str = name
a__ : Tuple = con
a__ : List[Any] = batch_size if batch_size else config.DEFAULT_MAX_BATCH_SIZE
a__ : Any = num_proc
a__ : Tuple = to_sql_kwargs
def _UpperCamelCase( self : List[Any] ):
a__ : Any = self.to_sql_kwargs.pop("sql" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("con" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("index" , lowerCamelCase__ )
a__ : int = self._write(index=lowerCamelCase__ , **self.to_sql_kwargs )
return written
def _UpperCamelCase( self : Any , lowerCamelCase__ : List[str] ):
a__, a__, a__ : Union[str, Any] = args
a__ : Any = {**to_sql_kwargs, "if_exists": "append"} if offset > 0 else to_sql_kwargs
a__ : Tuple = query_table(
table=self.dataset.data , key=slice(lowerCamelCase__ , offset + self.batch_size ) , indices=self.dataset._indices , )
a__ : str = batch.to_pandas()
a__ : List[Any] = df.to_sql(self.name , self.con , index=lowerCamelCase__ , **lowerCamelCase__ )
return num_rows or len(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : str = 0
if self.num_proc is None or self.num_proc == 1:
for offset in logging.tqdm(
range(0 , len(self.dataset ) , self.batch_size ) , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += self._batch_sql((offset, index, to_sql_kwargs) )
else:
a__, a__ : List[str] = len(self.dataset ), self.batch_size
with multiprocessing.Pool(self.num_proc ) as pool:
for num_rows in logging.tqdm(
pool.imap(
self._batch_sql , [(offset, index, to_sql_kwargs) for offset in range(0 , lowerCamelCase__ , lowerCamelCase__ )] , ) , total=(num_rows // batch_size) + 1 if num_rows % batch_size else num_rows // batch_size , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += num_rows
return written
| 37 | 1 |
import tempfile
import torch
from diffusers import PNDMScheduler
from .test_schedulers import SchedulerCommonTest
class A__ ( A__ ):
"""simple docstring"""
_lowercase = (PNDMScheduler,)
_lowercase = (('num_inference_steps', 5_0),)
def _UpperCamelCase( self : int , **lowerCamelCase__ : str ):
a__ : Optional[int] = {
"num_train_timesteps": 1_000,
"beta_start": 0.0001,
"beta_end": 0.02,
"beta_schedule": "linear",
}
config.update(**lowerCamelCase__ )
return config
def _UpperCamelCase( self : str , lowerCamelCase__ : Any=0 , **lowerCamelCase__ : Tuple ):
a__ : List[str] = dict(self.forward_default_kwargs )
a__ : Any = kwargs.pop("num_inference_steps" , lowerCamelCase__ )
a__ : Union[str, Any] = self.dummy_sample
a__ : Optional[int] = 0.1 * sample
a__ : List[Any] = [residual + 0.2, residual + 0.15, residual + 0.1, residual + 0.05]
for scheduler_class in self.scheduler_classes:
a__ : List[Any] = self.get_scheduler_config(**lowerCamelCase__ )
a__ : str = scheduler_class(**lowerCamelCase__ )
scheduler.set_timesteps(lowerCamelCase__ )
# copy over dummy past residuals
a__ : Optional[Any] = dummy_past_residuals[:]
with tempfile.TemporaryDirectory() as tmpdirname:
scheduler.save_config(lowerCamelCase__ )
a__ : Tuple = scheduler_class.from_pretrained(lowerCamelCase__ )
new_scheduler.set_timesteps(lowerCamelCase__ )
# copy over dummy past residuals
a__ : Optional[Any] = dummy_past_residuals[:]
a__ : int = scheduler.step_prk(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
a__ : Optional[Any] = new_scheduler.step_prk(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
assert torch.sum(torch.abs(output - new_output ) ) < 1E-5, "Scheduler outputs are not identical"
a__ : Optional[Any] = scheduler.step_plms(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
a__ : Dict = new_scheduler.step_plms(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
assert torch.sum(torch.abs(output - new_output ) ) < 1E-5, "Scheduler outputs are not identical"
def _UpperCamelCase( self : Tuple ):
pass
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : List[Any]=0 , **lowerCamelCase__ : List[Any] ):
a__ : List[Any] = dict(self.forward_default_kwargs )
a__ : List[Any] = kwargs.pop("num_inference_steps" , lowerCamelCase__ )
a__ : int = self.dummy_sample
a__ : List[str] = 0.1 * sample
a__ : Union[str, Any] = [residual + 0.2, residual + 0.15, residual + 0.1, residual + 0.05]
for scheduler_class in self.scheduler_classes:
a__ : List[str] = self.get_scheduler_config()
a__ : List[str] = scheduler_class(**lowerCamelCase__ )
scheduler.set_timesteps(lowerCamelCase__ )
# copy over dummy past residuals (must be after setting timesteps)
a__ : Optional[int] = dummy_past_residuals[:]
with tempfile.TemporaryDirectory() as tmpdirname:
scheduler.save_config(lowerCamelCase__ )
a__ : Optional[Any] = scheduler_class.from_pretrained(lowerCamelCase__ )
# copy over dummy past residuals
new_scheduler.set_timesteps(lowerCamelCase__ )
# copy over dummy past residual (must be after setting timesteps)
a__ : Optional[Any] = dummy_past_residuals[:]
a__ : List[str] = scheduler.step_prk(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
a__ : List[Any] = new_scheduler.step_prk(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
assert torch.sum(torch.abs(output - new_output ) ) < 1E-5, "Scheduler outputs are not identical"
a__ : Union[str, Any] = scheduler.step_plms(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
a__ : str = new_scheduler.step_plms(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
assert torch.sum(torch.abs(output - new_output ) ) < 1E-5, "Scheduler outputs are not identical"
def _UpperCamelCase( self : int , **lowerCamelCase__ : Tuple ):
a__ : Union[str, Any] = self.scheduler_classes[0]
a__ : Dict = self.get_scheduler_config(**lowerCamelCase__ )
a__ : Optional[Any] = scheduler_class(**lowerCamelCase__ )
a__ : Any = 10
a__ : List[str] = self.dummy_model()
a__ : Tuple = self.dummy_sample_deter
scheduler.set_timesteps(lowerCamelCase__ )
for i, t in enumerate(scheduler.prk_timesteps ):
a__ : Tuple = model(lowerCamelCase__ , lowerCamelCase__ )
a__ : str = scheduler.step_prk(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ ).prev_sample
for i, t in enumerate(scheduler.plms_timesteps ):
a__ : str = model(lowerCamelCase__ , lowerCamelCase__ )
a__ : Optional[int] = scheduler.step_plms(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ ).prev_sample
return sample
def _UpperCamelCase( self : str ):
a__ : Optional[Any] = dict(self.forward_default_kwargs )
a__ : Tuple = kwargs.pop("num_inference_steps" , lowerCamelCase__ )
for scheduler_class in self.scheduler_classes:
a__ : int = self.get_scheduler_config()
a__ : List[Any] = scheduler_class(**lowerCamelCase__ )
a__ : Dict = self.dummy_sample
a__ : Tuple = 0.1 * sample
if num_inference_steps is not None and hasattr(lowerCamelCase__ , "set_timesteps" ):
scheduler.set_timesteps(lowerCamelCase__ )
elif num_inference_steps is not None and not hasattr(lowerCamelCase__ , "set_timesteps" ):
a__ : str = num_inference_steps
# copy over dummy past residuals (must be done after set_timesteps)
a__ : int = [residual + 0.2, residual + 0.15, residual + 0.1, residual + 0.05]
a__ : Dict = dummy_past_residuals[:]
a__ : str = scheduler.step_prk(lowerCamelCase__ , 0 , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
a__ : Union[str, Any] = scheduler.step_prk(lowerCamelCase__ , 1 , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
self.assertEqual(output_a.shape , sample.shape )
self.assertEqual(output_a.shape , output_a.shape )
a__ : Dict = scheduler.step_plms(lowerCamelCase__ , 0 , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
a__ : str = scheduler.step_plms(lowerCamelCase__ , 1 , lowerCamelCase__ , **lowerCamelCase__ ).prev_sample
self.assertEqual(output_a.shape , sample.shape )
self.assertEqual(output_a.shape , output_a.shape )
def _UpperCamelCase( self : Optional[int] ):
for timesteps in [100, 1_000]:
self.check_over_configs(num_train_timesteps=lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
for steps_offset in [0, 1]:
self.check_over_configs(steps_offset=lowerCamelCase__ )
a__ : str = self.scheduler_classes[0]
a__ : str = self.get_scheduler_config(steps_offset=1 )
a__ : Optional[Any] = scheduler_class(**lowerCamelCase__ )
scheduler.set_timesteps(10 )
assert torch.equal(
scheduler.timesteps , torch.LongTensor(
[901, 851, 851, 801, 801, 751, 751, 701, 701, 651, 651, 601, 601, 501, 401, 301, 201, 101, 1] ) , )
def _UpperCamelCase( self : List[str] ):
for beta_start, beta_end in zip([0.0001, 0.001] , [0.002, 0.02] ):
self.check_over_configs(beta_start=lowerCamelCase__ , beta_end=lowerCamelCase__ )
def _UpperCamelCase( self : List[str] ):
for schedule in ["linear", "squaredcos_cap_v2"]:
self.check_over_configs(beta_schedule=lowerCamelCase__ )
def _UpperCamelCase( self : Tuple ):
for prediction_type in ["epsilon", "v_prediction"]:
self.check_over_configs(prediction_type=lowerCamelCase__ )
def _UpperCamelCase( self : Tuple ):
for t in [1, 5, 10]:
self.check_over_forward(time_step=lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] ):
for t, num_inference_steps in zip([1, 5, 10] , [10, 50, 100] ):
self.check_over_forward(num_inference_steps=lowerCamelCase__ )
def _UpperCamelCase( self : List[str] ):
# earlier version of set_timesteps() caused an error indexing alpha's with inference steps as power of 3
a__ : Optional[int] = 27
for scheduler_class in self.scheduler_classes:
a__ : int = self.dummy_sample
a__ : Optional[int] = 0.1 * sample
a__ : str = self.get_scheduler_config()
a__ : List[str] = scheduler_class(**lowerCamelCase__ )
scheduler.set_timesteps(lowerCamelCase__ )
# before power of 3 fix, would error on first step, so we only need to do two
for i, t in enumerate(scheduler.prk_timesteps[:2] ):
a__ : Dict = scheduler.step_prk(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ ).prev_sample
def _UpperCamelCase( self : Optional[Any] ):
with self.assertRaises(lowerCamelCase__ ):
a__ : Union[str, Any] = self.scheduler_classes[0]
a__ : Optional[Any] = self.get_scheduler_config()
a__ : Union[str, Any] = scheduler_class(**lowerCamelCase__ )
scheduler.step_plms(self.dummy_sample , 1 , self.dummy_sample ).prev_sample
def _UpperCamelCase( self : Any ):
a__ : Union[str, Any] = self.full_loop()
a__ : str = torch.sum(torch.abs(lowerCamelCase__ ) )
a__ : Dict = torch.mean(torch.abs(lowerCamelCase__ ) )
assert abs(result_sum.item() - 198.1318 ) < 1E-2
assert abs(result_mean.item() - 0.2580 ) < 1E-3
def _UpperCamelCase( self : Tuple ):
a__ : Dict = self.full_loop(prediction_type="v_prediction" )
a__ : Optional[int] = torch.sum(torch.abs(lowerCamelCase__ ) )
a__ : int = torch.mean(torch.abs(lowerCamelCase__ ) )
assert abs(result_sum.item() - 67.3986 ) < 1E-2
assert abs(result_mean.item() - 0.0878 ) < 1E-3
def _UpperCamelCase( self : Tuple ):
# We specify different beta, so that the first alpha is 0.99
a__ : Tuple = self.full_loop(set_alpha_to_one=lowerCamelCase__ , beta_start=0.01 )
a__ : Tuple = torch.sum(torch.abs(lowerCamelCase__ ) )
a__ : Tuple = torch.mean(torch.abs(lowerCamelCase__ ) )
assert abs(result_sum.item() - 230.0399 ) < 1E-2
assert abs(result_mean.item() - 0.2995 ) < 1E-3
def _UpperCamelCase( self : List[Any] ):
# We specify different beta, so that the first alpha is 0.99
a__ : List[str] = self.full_loop(set_alpha_to_one=lowerCamelCase__ , beta_start=0.01 )
a__ : Optional[int] = torch.sum(torch.abs(lowerCamelCase__ ) )
a__ : Union[str, Any] = torch.mean(torch.abs(lowerCamelCase__ ) )
assert abs(result_sum.item() - 186.9482 ) < 1E-2
assert abs(result_mean.item() - 0.2434 ) < 1E-3
| 37 |
import math
from datetime import datetime, timedelta
def UpperCamelCase_ ( __a ) -> datetime:
a__ : Union[str, Any] = year % 19
a__ : List[str] = year % 4
a__ : str = year % 7
a__ : Any = math.floor(year / 100 )
a__ : List[str] = math.floor((13 + 8 * leap_day_inhibits) / 25 )
a__ : Optional[int] = leap_day_inhibits / 4
a__ : Union[str, Any] = (
15 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 30
a__ : Dict = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
a__ : Any = (19 * metonic_cycle + secular_moon_shift) % 30
# PHM -> Paschal Full Moon
a__ : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 29 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 19 )
elif days_to_add == 28 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 18 )
else:
return datetime(__a , 3 , 22 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
UpperCamelCase : Tuple = """will be""" if year > datetime.now().year else """was"""
print(f"""Easter in {year} {tense} {gauss_easter(year)}""")
| 37 | 1 |
import os
import tempfile
import unittest
import uuid
from pathlib import Path
from transformers.testing_utils import get_tests_dir, require_soundfile, require_torch, require_vision
from transformers.tools.agent_types import AgentAudio, AgentImage, AgentText
from transformers.utils import is_soundfile_availble, is_torch_available, is_vision_available
if is_torch_available():
import torch
if is_soundfile_availble():
import soundfile as sf
if is_vision_available():
from PIL import Image
def UpperCamelCase_ ( __a="" ) -> str:
a__ : Dict = tempfile.mkdtemp()
return os.path.join(__a , str(uuid.uuida() ) + suffix )
@require_soundfile
@require_torch
class A__ ( unittest.TestCase ):
"""simple docstring"""
def _UpperCamelCase( self : List[str] ):
a__ : int = torch.rand(12 , dtype=torch.floataa ) - 0.5
a__ : Tuple = AgentAudio(lowerCamelCase__ )
a__ : Tuple = str(agent_type.to_string() )
# Ensure that the tensor and the agent_type's tensor are the same
self.assertTrue(torch.allclose(lowerCamelCase__ , agent_type.to_raw() , atol=1E-4 ) )
del agent_type
# Ensure the path remains even after the object deletion
self.assertTrue(os.path.exists(lowerCamelCase__ ) )
# Ensure that the file contains the same value as the original tensor
a__, a__ : int = sf.read(lowerCamelCase__ )
self.assertTrue(torch.allclose(lowerCamelCase__ , torch.tensor(lowerCamelCase__ ) , atol=1E-4 ) )
def _UpperCamelCase( self : int ):
a__ : Union[str, Any] = torch.rand(12 , dtype=torch.floataa ) - 0.5
a__ : Any = get_new_path(suffix=".wav" )
sf.write(lowerCamelCase__ , lowerCamelCase__ , 16_000 )
a__ : int = AgentAudio(lowerCamelCase__ )
self.assertTrue(torch.allclose(lowerCamelCase__ , agent_type.to_raw() , atol=1E-4 ) )
self.assertEqual(agent_type.to_string() , lowerCamelCase__ )
@require_vision
@require_torch
class A__ ( unittest.TestCase ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
a__ : Union[str, Any] = torch.randint(0 , 256 , (64, 64, 3) )
a__ : Optional[int] = AgentImage(lowerCamelCase__ )
a__ : List[str] = str(agent_type.to_string() )
# Ensure that the tensor and the agent_type's tensor are the same
self.assertTrue(torch.allclose(lowerCamelCase__ , agent_type._tensor , atol=1E-4 ) )
self.assertIsInstance(agent_type.to_raw() , Image.Image )
# Ensure the path remains even after the object deletion
del agent_type
self.assertTrue(os.path.exists(lowerCamelCase__ ) )
def _UpperCamelCase( self : Optional[Any] ):
a__ : Union[str, Any] = Path(get_tests_dir("fixtures/tests_samples/COCO" ) ) / "000000039769.png"
a__ : Any = Image.open(lowerCamelCase__ )
a__ : Tuple = AgentImage(lowerCamelCase__ )
self.assertTrue(path.samefile(agent_type.to_string() ) )
self.assertTrue(image == agent_type.to_raw() )
# Ensure the path remains even after the object deletion
del agent_type
self.assertTrue(os.path.exists(lowerCamelCase__ ) )
def _UpperCamelCase( self : str ):
a__ : Any = Path(get_tests_dir("fixtures/tests_samples/COCO" ) ) / "000000039769.png"
a__ : Any = Image.open(lowerCamelCase__ )
a__ : Tuple = AgentImage(lowerCamelCase__ )
self.assertFalse(path.samefile(agent_type.to_string() ) )
self.assertTrue(image == agent_type.to_raw() )
# Ensure the path remains even after the object deletion
del agent_type
self.assertTrue(os.path.exists(lowerCamelCase__ ) )
class A__ ( unittest.TestCase ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
a__ : Tuple = "Hey!"
a__ : Tuple = AgentText(lowerCamelCase__ )
self.assertEqual(lowerCamelCase__ , agent_type.to_string() )
self.assertEqual(lowerCamelCase__ , agent_type.to_raw() )
self.assertEqual(lowerCamelCase__ , lowerCamelCase__ )
| 37 |
import gc
import importlib.metadata
import tempfile
import unittest
from packaging import version
from transformers import (
AutoModel,
AutoModelForCausalLM,
AutoModelForSeqaSeqLM,
AutoModelForSequenceClassification,
AutoTokenizer,
BitsAndBytesConfig,
pipeline,
)
from transformers.testing_utils import (
is_torch_available,
require_accelerate,
require_bitsandbytes,
require_torch,
require_torch_gpu,
require_torch_multi_gpu,
slow,
)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
if model.config.model_type == "gpt2":
return model.transformer.h[0].mlp.c_fc
return model.transformer.h[0].mlp.dense_ah_to_h
if is_torch_available():
import torch
import torch.nn as nn
class A__ ( nn.Module ):
"""simple docstring"""
def __init__( self : List[str] , lowerCamelCase__ : nn.Module , lowerCamelCase__ : int ):
super().__init__()
a__ : int = module
a__ : Any = nn.Sequential(
nn.Linear(module.in_features , lowerCamelCase__ , bias=lowerCamelCase__ ) , nn.Linear(lowerCamelCase__ , module.out_features , bias=lowerCamelCase__ ) , )
a__ : Tuple = (2.0 / (5 * min(module.in_features , module.out_features ))) ** 0.5
nn.init.normal_(self.adapter[0].weight , std=lowerCamelCase__ )
nn.init.zeros_(self.adapter[1].weight )
self.adapter.to(module.weight.device )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Optional[int] , *lowerCamelCase__ : int , **lowerCamelCase__ : Dict ):
return self.module(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ ) + self.adapter(lowerCamelCase__ )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
_lowercase = 'bigscience/bloom-1b7'
# Constant values
_lowercase = 2.1_09_65_95_52_69_25_74
_lowercase = 'Hello my name is'
_lowercase = set()
EXPECTED_OUTPUTS.add('Hello my name is John and I am a professional photographer. I' )
EXPECTED_OUTPUTS.add('Hello my name is John.\nI am a friend of your father.\n' )
EXPECTED_OUTPUTS.add('Hello my name is John Doe, I am a student at the University' )
_lowercase = 1_0
def _UpperCamelCase( self : Dict ):
# Models and tokenizer
a__ : List[str] = AutoTokenizer.from_pretrained(self.model_name )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Union[str, Any] ):
super().setUp()
# Models and tokenizer
a__ : List[Any] = AutoModelForCausalLM.from_pretrained(
self.model_name , torch_dtype=torch.floataa , device_map="auto" )
a__ : Union[str, Any] = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.model_fpaa
del self.model_abit
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : List[Any] ):
a__ : str = self.model_abit.config
self.assertTrue(hasattr(lowerCamelCase__ , "quantization_config" ) )
a__ : Optional[Any] = config.to_dict()
a__ : int = config.to_diff_dict()
a__ : List[str] = config.to_json_string()
def _UpperCamelCase( self : int ):
from bitsandbytes.nn import Paramsabit
a__ : List[Any] = self.model_fpaa.get_memory_footprint()
a__ : str = self.model_abit.get_memory_footprint()
self.assertAlmostEqual(mem_fpaa / mem_abit , self.EXPECTED_RELATIVE_DIFFERENCE )
a__ : Optional[Any] = get_some_linear_layer(self.model_abit )
self.assertTrue(linear.weight.__class__ == Paramsabit )
def _UpperCamelCase( self : Tuple ):
from transformers import TaPreTrainedModel
self.model_fpaa.get_memory_footprint()
self.model_abit.get_memory_footprint()
for name, module in self.model_abit.named_modules():
if isinstance(lowerCamelCase__ , torch.nn.Linear ):
if name not in ["lm_head"] + TaPreTrainedModel._keep_in_fpaa_modules:
# 4-bit parameters are packed in uint8 variables
self.assertTrue(module.weight.dtype == torch.uinta )
def _UpperCamelCase( self : str ):
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Tuple = self.model_abit.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[Any] = BitsAndBytesConfig()
a__ : Optional[int] = True
a__ : int = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , device_map="auto" )
a__ : str = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : int = model_abit_from_config.generate(
input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : Dict ):
with self.assertRaises(lowerCamelCase__ ), tempfile.TemporaryDirectory() as tmpdirname:
self.model_abit.save_pretrained(lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : int = BitsAndBytesConfig()
with self.assertRaises(lowerCamelCase__ ):
a__ : Dict = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , load_in_abit=lowerCamelCase__ , device_map="auto" , bnb_abit_quant_type="nf4" , )
def _UpperCamelCase( self : int ):
with self.assertRaises(lowerCamelCase__ ):
# Tries with `str`
self.model_abit.to("cpu" )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `dtype``
self.model_abit.to(torch.floataa )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.to(torch.device("cuda:0" ) )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.float()
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.half()
# Test if we did not break anything
a__ : int = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Any = self.model_fpaa.to(torch.floataa )
a__ : List[Any] = self.model_fpaa.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.to("cpu" )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.half()
# Check this does not throw an error
a__ : Dict = self.model_fpaa.float()
def _UpperCamelCase( self : Dict ):
a__ : List[str] = AutoModelForSeqaSeqLM.from_pretrained("t5-small" , load_in_abit=lowerCamelCase__ , device_map="auto" )
self.assertTrue(model.decoder.block[0].layer[2].DenseReluDense.wo.weight.dtype == torch.floataa )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
@classmethod
def _UpperCamelCase( cls : str ):
a__ : Dict = "t5-small"
a__ : List[Any] = "google/flan-t5-small" # flan-t5 uses dense-act instead of dense-relu-dense
a__ : int = AutoTokenizer.from_pretrained(cls.model_name )
a__ : str = "Translate in German: Hello, my dog is cute"
def _UpperCamelCase( self : Optional[int] ):
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Optional[int] ):
from transformers import TaForConditionalGeneration
a__ : List[Any] = TaForConditionalGeneration._keep_in_fpaa_modules
a__ : Optional[Any] = None
# test with `t5-small`
a__ : Any = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : Dict = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Any = model.generate(**lowerCamelCase__ )
a__ : Union[str, Any] = modules
def _UpperCamelCase( self : List[Any] ):
import bitsandbytes as bnb
from transformers import TaForConditionalGeneration
# test with `t5-small`
a__ : List[str] = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# there was a bug with decoders - this test checks that it is fixed
self.assertTrue(isinstance(model.decoder.block[0].layer[0].SelfAttention.q , bnb.nn.Linearabit ) )
a__ : Union[str, Any] = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : int = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Any = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Optional[int] = model.generate(**lowerCamelCase__ )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : List[str] ):
super().setUp()
# model_name
a__ : Union[str, Any] = "bigscience/bloom-560m"
a__ : Union[str, Any] = "t5-small"
# Different types of model
a__ : int = AutoModel.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Sequence classification model
a__ : Dict = AutoModelForSequenceClassification.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# CausalLM model
a__ : str = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Seq2seq model
a__ : Dict = AutoModelForSeqaSeqLM.from_pretrained(
self.seq_to_seq_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.base_model
del self.sequence_model
del self.model_abit
del self.seq_to_seq_model
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Union[str, Any] ):
from bitsandbytes.nn import Paramsabit
self.assertTrue(self.base_model.h[-1].mlp.dense_ah_to_h.weight.__class__ == Paramsabit )
# Other heads should be nn.Parameter
self.assertTrue(self.model_abit.lm_head.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.sequence_model.score.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.seq_to_seq_model.lm_head.weight.__class__ == torch.nn.Parameter )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
super().setUp()
def _UpperCamelCase( self : int ):
del self.pipe
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Tuple ):
a__ : int = pipeline(
"text-generation" , model=self.model_name , model_kwargs={"device_map": "auto", "load_in_4bit": True, "torch_dtype": torch.floataa} , max_new_tokens=self.MAX_NEW_TOKENS , )
# Real second forward pass
a__ : Tuple = self.pipe(self.input_text )
self.assertIn(pipeline_output[0]["generated_text"] , self.EXPECTED_OUTPUTS )
@require_torch_multi_gpu
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Tuple ):
super().setUp()
def _UpperCamelCase( self : List[Any] ):
a__ : str = AutoModelForCausalLM.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="balanced" )
# Check correct device map
self.assertEqual(set(model_parallel.hf_device_map.values() ) , {0, 1} )
# Check that inference pass works on the model
a__ : List[Any] = self.tokenizer(self.input_text , return_tensors="pt" )
# Second real batch
a__ : List[Any] = model_parallel.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_parallel[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
a__ : Any = "facebook/opt-350m"
super().setUp()
def _UpperCamelCase( self : int ):
if version.parse(importlib.metadata.version("bitsandbytes" ) ) < version.parse("0.37.0" ):
return
# Step 1: freeze all parameters
a__ : Tuple = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ )
self.assertEqual(set(model.hf_device_map.values() ) , {torch.cuda.current_device()} )
for param in model.parameters():
a__ : Any = False # freeze the model - train adapters later
if param.ndim == 1:
# cast the small parameters (e.g. layernorm) to fp32 for stability
a__ : Tuple = param.data.to(torch.floataa )
# Step 2: add adapters
for _, module in model.named_modules():
if "OPTAttention" in repr(type(lowerCamelCase__ ) ):
a__ : Dict = LoRALayer(module.q_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.k_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.v_proj , rank=16 )
# Step 3: dummy batch
a__ : Dict = self.tokenizer("Test batch " , return_tensors="pt" ).to(0 )
# Step 4: Check if the gradient is not None
with torch.cuda.amp.autocast():
a__ : Optional[Any] = model.forward(**lowerCamelCase__ )
out.logits.norm().backward()
for module in model.modules():
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertTrue(module.adapter[1].weight.grad is not None )
self.assertTrue(module.adapter[1].weight.grad.norm().item() > 0 )
elif isinstance(lowerCamelCase__ , nn.Embedding ):
self.assertTrue(module.weight.grad is None )
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gpt2-xl'
_lowercase = 3.31_91_85_48_54_15_21_87
| 37 | 1 |
import re
from pathlib import Path
from unittest import TestCase
import pytest
@pytest.mark.integration
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : str ):
with open(lowerCamelCase__ , encoding="utf-8" ) as input_file:
a__ : Dict = re.compile(r"(?!.*\b(?:encoding|rb|w|wb|w+|wb+|ab|ab+)\b)(?<=\s)(open)\((.*)\)" )
a__ : Any = input_file.read()
a__ : Tuple = regexp.search(lowerCamelCase__ )
return match
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : str ):
with open(lowerCamelCase__ , encoding="utf-8" ) as input_file:
a__ : str = re.compile(r"#[^\r\n]*print\(|\"[^\r\n]*print\(|\"\"\".*?print\(.*?\"\"\"|(print\()" , re.DOTALL )
a__ : List[Any] = input_file.read()
# use `re.finditer` to handle the case where the ignored groups would be matched first by `re.search`
a__ : str = regexp.finditer(lowerCamelCase__ )
a__ : Any = [match for match in matches if match is not None and match.group(1 ) is not None]
return matches[0] if matches else None
def _UpperCamelCase( self : int ):
a__ : Optional[int] = Path("./datasets" )
a__ : Dict = list(dataset_paths.absolute().glob("**/*.py" ) )
for dataset in dataset_files:
if self._no_encoding_on_file_open(str(lowerCamelCase__ ) ):
raise AssertionError(f'''open(...) must use utf-8 encoding in {dataset}''' )
def _UpperCamelCase( self : Optional[int] ):
a__ : Tuple = Path("./datasets" )
a__ : Union[str, Any] = list(dataset_paths.absolute().glob("**/*.py" ) )
for dataset in dataset_files:
if self._no_print_statements(str(lowerCamelCase__ ) ):
raise AssertionError(f'''print statement found in {dataset}. Use datasets.logger/logging instead.''' )
| 37 |
import inspect
import unittest
from datasets import load_dataset
from packaging import version
from transformers import BeitConfig
from transformers.models.auto import get_values
from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_vision_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, _config_zero_init, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import (
MODEL_MAPPING,
BeitForImageClassification,
BeitForMaskedImageModeling,
BeitForSemanticSegmentation,
BeitModel,
)
from transformers.models.beit.modeling_beit import BEIT_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
import PIL
from PIL import Image
from transformers import BeitImageProcessor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Optional[int]=100 , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[int]=30 , lowerCamelCase__ : Union[str, Any]=2 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : List[Any]=True , lowerCamelCase__ : Tuple=True , lowerCamelCase__ : int=32 , lowerCamelCase__ : Union[str, Any]=4 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Union[str, Any]=37 , lowerCamelCase__ : List[Any]="gelu" , lowerCamelCase__ : List[Any]=0.1 , lowerCamelCase__ : int=0.1 , lowerCamelCase__ : Union[str, Any]=10 , lowerCamelCase__ : str=0.02 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]=[0, 1, 2, 3] , ):
a__ : Dict = parent
a__ : Dict = 100
a__ : Optional[int] = batch_size
a__ : Union[str, Any] = image_size
a__ : Any = patch_size
a__ : Optional[Any] = num_channels
a__ : int = is_training
a__ : List[str] = use_labels
a__ : Optional[Any] = hidden_size
a__ : List[Any] = num_hidden_layers
a__ : str = num_attention_heads
a__ : str = intermediate_size
a__ : int = hidden_act
a__ : List[Any] = hidden_dropout_prob
a__ : Dict = attention_probs_dropout_prob
a__ : Union[str, Any] = type_sequence_label_size
a__ : Optional[Any] = initializer_range
a__ : List[str] = scope
a__ : int = out_indices
a__ : List[str] = num_labels
# in BeiT, the seq length equals the number of patches + 1 (we add 1 for the [CLS] token)
a__ : Optional[int] = (image_size // patch_size) ** 2
a__ : Union[str, Any] = num_patches + 1
def _UpperCamelCase( self : int ):
a__ : Optional[int] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
a__ : Optional[Any] = None
a__ : Tuple = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : Dict = ids_tensor([self.batch_size, self.image_size, self.image_size] , self.num_labels )
a__ : Optional[int] = self.get_config()
return config, pixel_values, labels, pixel_labels
def _UpperCamelCase( self : Tuple ):
return BeitConfig(
vocab_size=self.vocab_size , image_size=self.image_size , patch_size=self.patch_size , num_channels=self.num_channels , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , out_indices=self.out_indices , )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : int , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any ):
a__ : str = BeitModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[str] , lowerCamelCase__ : Any , lowerCamelCase__ : List[str] , lowerCamelCase__ : Tuple ):
a__ : int = BeitForMaskedImageModeling(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[Any] = model(lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length - 1, self.vocab_size) )
def _UpperCamelCase( self : str , lowerCamelCase__ : Any , lowerCamelCase__ : str , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Any ):
a__ : List[str] = self.type_sequence_label_size
a__ : Optional[Any] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
# test greyscale images
a__ : Optional[Any] = 1
a__ : List[str] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = floats_tensor([self.batch_size, 1, self.image_size, self.image_size] )
a__ : Union[str, Any] = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def _UpperCamelCase( self : Any , lowerCamelCase__ : str , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
a__ : int = self.num_labels
a__ : List[str] = BeitForSemanticSegmentation(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Tuple = model(lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
def _UpperCamelCase( self : Optional[int] ):
a__ : Any = self.prepare_config_and_inputs()
a__, a__, a__, a__ : Union[str, Any] = config_and_inputs
a__ : Dict = {"pixel_values": pixel_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(BeitModel, BeitForImageClassification, BeitForMaskedImageModeling, BeitForSemanticSegmentation)
if is_torch_available()
else ()
)
_lowercase = (
{
'feature-extraction': BeitModel,
'image-classification': BeitForImageClassification,
'image-segmentation': BeitForSemanticSegmentation,
}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Any ):
a__ : int = BeitModelTester(self )
a__ : Optional[Any] = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[Any] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="BEiT does not use inputs_embeds" )
def _UpperCamelCase( self : str ):
pass
@require_torch_multi_gpu
@unittest.skip(reason="BEiT has some layers using `add_module` which doesn't work well with `nn.DataParallel`" )
def _UpperCamelCase( self : Dict ):
pass
def _UpperCamelCase( self : Optional[Any] ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : List[str] = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Optional[int] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : str ):
a__, a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : int = model_class(lowerCamelCase__ )
a__ : str = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Any = ["pixel_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
def _UpperCamelCase( self : int ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_lm(*lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] ):
a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_semantic_segmentation(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
if not self.model_tester.is_training:
return
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : str = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]:
continue
a__ : List[str] = model_class(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.train()
a__ : Any = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : Tuple = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : Tuple ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
if not self.model_tester.is_training:
return
a__ : List[Any] = False
a__ : List[str] = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if (
model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]
or not model_class.supports_gradient_checkpointing
):
continue
a__ : Optional[Any] = model_class(lowerCamelCase__ )
model.gradient_checkpointing_enable()
model.to(lowerCamelCase__ )
model.train()
a__ : Union[str, Any] = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : int = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : List[str] ):
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : Dict = _config_zero_init(lowerCamelCase__ )
for model_class in self.all_model_classes:
a__ : str = model_class(config=lowerCamelCase__ )
for name, param in model.named_parameters():
# we skip lambda parameters as these require special initial values
# determined by config.layer_scale_init_value
if "lambda" in name:
continue
if param.requires_grad:
self.assertIn(
((param.data.mean() * 1E9).round() / 1E9).item() , [0.0, 1.0] , msg=f'''Parameter {name} of model {model_class} seems not properly initialized''' , )
@slow
def _UpperCamelCase( self : Optional[int] ):
for model_name in BEIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Tuple = BeitModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : List[Any] = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
return image
@require_torch
@require_vision
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : Optional[int] ):
return BeitImageProcessor.from_pretrained("microsoft/beit-base-patch16-224" ) if is_vision_available() else None
@slow
def _UpperCamelCase( self : str ):
a__ : int = BeitForMaskedImageModeling.from_pretrained("microsoft/beit-base-patch16-224-pt22k" ).to(lowerCamelCase__ )
a__ : Optional[Any] = self.default_image_processor
a__ : Dict = prepare_img()
a__ : Optional[int] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).pixel_values.to(lowerCamelCase__ )
# prepare bool_masked_pos
a__ : Optional[Any] = torch.ones((1, 196) , dtype=torch.bool ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(pixel_values=lowerCamelCase__ , bool_masked_pos=lowerCamelCase__ )
a__ : Tuple = outputs.logits
# verify the logits
a__ : List[str] = torch.Size((1, 196, 8_192) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[int] = torch.tensor(
[[-3.2437, 0.5072, -13.9174], [-3.2456, 0.4948, -13.9401], [-3.2033, 0.5121, -13.8550]] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[bool_masked_pos][:3, :3] , lowerCamelCase__ , atol=1E-2 ) )
@slow
def _UpperCamelCase( self : Dict ):
a__ : str = BeitForImageClassification.from_pretrained("microsoft/beit-base-patch16-224" ).to(lowerCamelCase__ )
a__ : int = self.default_image_processor
a__ : List[Any] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Union[str, Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 1_000) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = torch.tensor([-1.2385, -1.0987, -1.0108] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Tuple = 281
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : Any ):
a__ : Dict = BeitForImageClassification.from_pretrained("microsoft/beit-large-patch16-224-pt22k-ft22k" ).to(
lowerCamelCase__ )
a__ : str = self.default_image_processor
a__ : List[str] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Dict = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Optional[int] = torch.Size((1, 21_841) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[Any] = torch.tensor([1.6881, -0.2787, 0.5901] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Optional[Any] = 2_396
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
a__ : Optional[Any] = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : Tuple = model.to(lowerCamelCase__ )
a__ : List[Any] = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Tuple = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : Union[str, Any] = Image.open(ds[0]["file"] )
a__ : List[Any] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Optional[Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Tuple = torch.Size((1, 150, 160, 160) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = version.parse(PIL.__version__ ) < version.parse("9.0.0" )
if is_pillow_less_than_a:
a__ : Dict = torch.tensor(
[
[[-4.9225, -2.3954, -3.0522], [-2.8822, -1.0046, -1.7561], [-2.9549, -1.3228, -2.1347]],
[[-5.8168, -3.4129, -4.0778], [-3.8651, -2.2214, -3.0277], [-3.8356, -2.4643, -3.3535]],
[[-0.0078, 3.9952, 4.0754], [2.9856, 4.6944, 5.0035], [3.2413, 4.7813, 4.9969]],
] , device=lowerCamelCase__ , )
else:
a__ : Dict = torch.tensor(
[
[[-4.8960, -2.3688, -3.0355], [-2.8478, -0.9836, -1.7418], [-2.9449, -1.3332, -2.1456]],
[[-5.8081, -3.4124, -4.1006], [-3.8561, -2.2081, -3.0323], [-3.8365, -2.4601, -3.3669]],
[[-0.0309, 3.9868, 4.0540], [2.9640, 4.6877, 4.9976], [3.2081, 4.7690, 4.9942]],
] , device=lowerCamelCase__ , )
self.assertTrue(torch.allclose(logits[0, :3, :3, :3] , lowerCamelCase__ , atol=1E-4 ) )
@slow
def _UpperCamelCase( self : Tuple ):
a__ : str = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : List[Any] = model.to(lowerCamelCase__ )
a__ : int = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Optional[int] = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : str = Image.open(ds[0]["file"] )
a__ : str = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : List[Any] = model(**lowerCamelCase__ )
a__ : Any = outputs.logits.detach().cpu()
a__ : List[Any] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ , target_sizes=[(500, 300)] )
a__ : Optional[int] = torch.Size((500, 300) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
a__ : List[str] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ )
a__ : Any = torch.Size((160, 160) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
| 37 | 1 |
from __future__ import annotations
def UpperCamelCase_ ( __a , __a = None , __a = None ) -> None:
if start is None:
a__ : int = 0
if end is None:
a__ : str = len(__a ) - 1
if start >= end:
return
a__ : Union[str, Any] = (start + end) // 2
slowsort(__a , __a , __a )
slowsort(__a , mid + 1 , __a )
if sequence[end] < sequence[mid]:
a__, a__ : int = sequence[mid], sequence[end]
slowsort(__a , __a , end - 1 )
if __name__ == "__main__":
from doctest import testmod
testmod()
| 37 |
import re
import jax.numpy as jnp
from flax.traverse_util import flatten_dict, unflatten_dict
from jax.random import PRNGKey
from ..utils import logging
UpperCamelCase : Dict = logging.get_logger(__name__)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
a__ : Tuple = R"\w+[.]\d+"
a__ : List[Any] = re.findall(__a , __a )
for pat in pats:
a__ : Union[str, Any] = key.replace(__a , "_".join(pat.split("." ) ) )
return key
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : List[str] = pt_tuple_key[:-1] + ("scale",)
if (
any("norm" in str_ for str_ in pt_tuple_key )
and (pt_tuple_key[-1] == "bias")
and (pt_tuple_key[:-1] + ("bias",) not in random_flax_state_dict)
and (pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict)
):
a__ : Any = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
elif pt_tuple_key[-1] in ["weight", "gamma"] and pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict:
a__ : Optional[Any] = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
# embedding
if pt_tuple_key[-1] == "weight" and pt_tuple_key[:-1] + ("embedding",) in random_flax_state_dict:
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("embedding",)
return renamed_pt_tuple_key, pt_tensor
# conv layer
a__ : List[str] = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight" and pt_tensor.ndim == 4:
a__ : str = pt_tensor.transpose(2 , 3 , 1 , 0 )
return renamed_pt_tuple_key, pt_tensor
# linear layer
a__ : Tuple = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight":
a__ : Tuple = pt_tensor.T
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm weight
a__ : Optional[Any] = pt_tuple_key[:-1] + ("weight",)
if pt_tuple_key[-1] == "gamma":
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm bias
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("bias",)
if pt_tuple_key[-1] == "beta":
return renamed_pt_tuple_key, pt_tensor
return pt_tuple_key, pt_tensor
def UpperCamelCase_ ( __a , __a , __a=42 ) -> str:
# Step 1: Convert pytorch tensor to numpy
a__ : Optional[int] = {k: v.numpy() for k, v in pt_state_dict.items()}
# Step 2: Since the model is stateless, get random Flax params
a__ : Tuple = flax_model.init_weights(PRNGKey(__a ) )
a__ : Optional[Any] = flatten_dict(__a )
a__ : Union[str, Any] = {}
# Need to change some parameters name to match Flax names
for pt_key, pt_tensor in pt_state_dict.items():
a__ : Optional[int] = rename_key(__a )
a__ : Optional[int] = tuple(renamed_pt_key.split("." ) )
# Correctly rename weight parameters
a__, a__ : Union[str, Any] = rename_key_and_reshape_tensor(__a , __a , __a )
if flax_key in random_flax_state_dict:
if flax_tensor.shape != random_flax_state_dict[flax_key].shape:
raise ValueError(
f'''PyTorch checkpoint seems to be incorrect. Weight {pt_key} was expected to be of shape '''
f'''{random_flax_state_dict[flax_key].shape}, but is {flax_tensor.shape}.''' )
# also add unexpected weight so that warning is thrown
a__ : str = jnp.asarray(__a )
return unflatten_dict(__a )
| 37 | 1 |
import dataclasses
import json
import warnings
from dataclasses import dataclass, field
from time import time
from typing import List
from ..utils import logging
UpperCamelCase : Dict = logging.get_logger(__name__)
def UpperCamelCase_ ( __a=None , __a=None ) -> List[str]:
return field(default_factory=lambda: default , metadata=__a )
@dataclass
class A__ :
"""simple docstring"""
_lowercase = list_field(
default=[] , metadata={
'help': (
'Model checkpoints to be provided to the AutoModel classes. Leave blank to benchmark the base version'
' of all available models'
)
} , )
_lowercase = list_field(
default=[8] , metadata={'help': 'List of batch sizes for which memory and time performance will be evaluated'} )
_lowercase = list_field(
default=[8, 3_2, 1_2_8, 5_1_2] , metadata={'help': 'List of sequence lengths for which memory and time performance will be evaluated'} , )
_lowercase = field(
default=A__ , metadata={'help': 'Whether to benchmark inference of model. Inference can be disabled via --no-inference.'} , )
_lowercase = field(
default=A__ , metadata={'help': 'Whether to run on available cuda devices. Cuda can be disabled via --no-cuda.'} , )
_lowercase = field(
default=A__ , metadata={'help': 'Whether to run on available tpu devices. TPU can be disabled via --no-tpu.'} )
_lowercase = field(default=A__ , metadata={'help': 'Use FP16 to accelerate inference.'} )
_lowercase = field(default=A__ , metadata={'help': 'Benchmark training of model'} )
_lowercase = field(default=A__ , metadata={'help': 'Verbose memory tracing'} )
_lowercase = field(
default=A__ , metadata={'help': 'Whether to perform speed measurements. Speed measurements can be disabled via --no-speed.'} , )
_lowercase = field(
default=A__ , metadata={
'help': 'Whether to perform memory measurements. Memory measurements can be disabled via --no-memory'
} , )
_lowercase = field(default=A__ , metadata={'help': 'Trace memory line by line'} )
_lowercase = field(default=A__ , metadata={'help': 'Save result to a CSV file'} )
_lowercase = field(default=A__ , metadata={'help': 'Save all print statements in a log file'} )
_lowercase = field(default=A__ , metadata={'help': 'Whether to print environment information'} )
_lowercase = field(
default=A__ , metadata={
'help': (
'Whether to use multiprocessing for memory and speed measurement. It is highly recommended to use'
' multiprocessing for accurate CPU and GPU memory measurements. This option should only be disabled'
' for debugging / testing and on TPU.'
)
} , )
_lowercase = field(
default=F'inference_time_{round(time() )}.csv' , metadata={'help': 'CSV filename used if saving time results to csv.'} , )
_lowercase = field(
default=F'inference_memory_{round(time() )}.csv' , metadata={'help': 'CSV filename used if saving memory results to csv.'} , )
_lowercase = field(
default=F'train_time_{round(time() )}.csv' , metadata={'help': 'CSV filename used if saving time results to csv for training.'} , )
_lowercase = field(
default=F'train_memory_{round(time() )}.csv' , metadata={'help': 'CSV filename used if saving memory results to csv for training.'} , )
_lowercase = field(
default=F'env_info_{round(time() )}.csv' , metadata={'help': 'CSV filename used if saving environment information.'} , )
_lowercase = field(
default=F'log_{round(time() )}.csv' , metadata={'help': 'Log filename used if print statements are saved in log.'} , )
_lowercase = field(default=3 , metadata={'help': 'Times an experiment will be run.'} )
_lowercase = field(
default=A__ , metadata={
'help': (
'Instead of loading the model as defined in `config.architectures` if exists, just load the pretrain'
' model weights.'
)
} , )
def _UpperCamelCase( self : Optional[int] ):
warnings.warn(
f'''The class {self.__class__} is deprecated. Hugging Face Benchmarking utils'''
" are deprecated in general and it is advised to use external Benchmarking libraries "
" to benchmark Transformer models." , lowerCamelCase__ , )
def _UpperCamelCase( self : Union[str, Any] ):
return json.dumps(dataclasses.asdict(self ) , indent=2 )
@property
def _UpperCamelCase( self : List[str] ):
if len(self.models ) <= 0:
raise ValueError(
"Please make sure you provide at least one model name / model identifier, *e.g.* `--models"
" bert-base-cased` or `args.models = ['bert-base-cased']." )
return self.models
@property
def _UpperCamelCase( self : Any ):
if not self.multi_process:
return False
elif self.is_tpu:
logger.info("Multiprocessing is currently not possible on TPU." )
return False
else:
return True
| 37 |
from transformers import HfArgumentParser, TensorFlowBenchmark, TensorFlowBenchmarkArguments
def UpperCamelCase_ ( ) -> int:
a__ : Any = HfArgumentParser(__a )
a__ : Any = parser.parse_args_into_dataclasses()[0]
a__ : Optional[int] = TensorFlowBenchmark(args=__a )
try:
a__ : Optional[int] = parser.parse_args_into_dataclasses()[0]
except ValueError as e:
a__ : Tuple = "Arg --no_{0} is no longer used, please use --no-{0} instead."
a__ : List[Any] = " ".join(str(__a ).split(" " )[:-1] )
a__ : str = ""
a__ : List[Any] = eval(str(__a ).split(" " )[-1] )
a__ : List[str] = []
for arg in depreciated_args:
# arg[2:] removes '--'
if arg[2:] in TensorFlowBenchmark.deprecated_args:
# arg[5:] removes '--no_'
full_error_msg += arg_error_msg.format(arg[5:] )
else:
wrong_args.append(__a )
if len(__a ) > 0:
a__ : Tuple = full_error_msg + begin_error_msg + str(__a )
raise ValueError(__a )
benchmark.run()
if __name__ == "__main__":
main()
| 37 | 1 |
# Copyright 2023 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import TYPE_CHECKING
from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available
UpperCamelCase : Any = {
"""configuration_xmod""": [
"""XMOD_PRETRAINED_CONFIG_ARCHIVE_MAP""",
"""XmodConfig""",
"""XmodOnnxConfig""",
],
}
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
UpperCamelCase : str = [
"""XMOD_PRETRAINED_MODEL_ARCHIVE_LIST""",
"""XmodForCausalLM""",
"""XmodForMaskedLM""",
"""XmodForMultipleChoice""",
"""XmodForQuestionAnswering""",
"""XmodForSequenceClassification""",
"""XmodForTokenClassification""",
"""XmodModel""",
"""XmodPreTrainedModel""",
]
if TYPE_CHECKING:
from .configuration_xmod import XMOD_PRETRAINED_CONFIG_ARCHIVE_MAP, XmodConfig, XmodOnnxConfig
try:
if not is_torch_available():
raise OptionalDependencyNotAvailable()
except OptionalDependencyNotAvailable:
pass
else:
from .modeling_xmod import (
XMOD_PRETRAINED_MODEL_ARCHIVE_LIST,
XmodForCausalLM,
XmodForMaskedLM,
XmodForMultipleChoice,
XmodForQuestionAnswering,
XmodForSequenceClassification,
XmodForTokenClassification,
XmodModel,
XmodPreTrainedModel,
)
else:
import sys
UpperCamelCase : Any = _LazyModule(__name__, globals()["""__file__"""], _import_structure, module_spec=__spec__)
| 37 |
import argparse
import ast
import logging
import os
import sys
import pandas as pd
import torch
from tqdm import tqdm
from transformers import BartForConditionalGeneration, RagRetriever, RagSequenceForGeneration, RagTokenForGeneration
from transformers import logging as transformers_logging
sys.path.append(os.path.join(os.getcwd())) # noqa: E402 # isort:skip
from utils_rag import exact_match_score, fa_score # noqa: E402 # isort:skip
UpperCamelCase : Optional[int] = logging.getLogger(__name__)
logging.basicConfig(level=logging.INFO)
transformers_logging.set_verbosity_info()
def UpperCamelCase_ ( __a ) -> Any:
if "token" in model_name_or_path:
return "rag_token"
if "sequence" in model_name_or_path:
return "rag_sequence"
if "bart" in model_name_or_path:
return "bart"
return None
def UpperCamelCase_ ( __a , __a , __a ) -> Any:
return max(metric_fn(__a , __a ) for gt in ground_truths )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = []
if args.gold_data_mode == "qa":
a__ : Any = pd.read_csv(__a , sep="\t" , header=__a )
for answer_list in data[1]:
a__ : Union[str, Any] = ast.literal_eval(__a )
answers.append(__a )
else:
a__ : List[str] = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : List[str] = [[reference] for reference in references]
a__ : List[str] = 0
for prediction, ground_truths in zip(__a , __a ):
total += 1
em += metric_max_over_ground_truths(__a , __a , __a )
fa += metric_max_over_ground_truths(__a , __a , __a )
a__ : Dict = 100.0 * em / total
a__ : Optional[Any] = 100.0 * fa / total
logger.info(f'''F1: {fa:.2f}''' )
logger.info(f'''EM: {em:.2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = args.k
a__ : str = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = 0
for hypo, reference in zip(__a , __a ):
a__ : Any = set(hypo.split("\t" )[:k] )
a__ : Union[str, Any] = set(reference.split("\t" ) )
total += 1
em += len(hypo_provenance & ref_provenance ) / k
a__ : Union[str, Any] = 100.0 * em / total
logger.info(f'''Precision@{k}: {em: .2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
def strip_title(__a ):
if title.startswith("\"" ):
a__ : Optional[Any] = title[1:]
if title.endswith("\"" ):
a__ : Union[str, Any] = title[:-1]
return title
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a , )["input_ids"].to(args.device )
a__ : Optional[int] = rag_model.rag.question_encoder(__a )
a__ : Union[str, Any] = question_enc_outputs[0]
a__ : Optional[int] = rag_model.retriever(
__a , question_enc_pool_output.cpu().detach().to(torch.floataa ).numpy() , prefix=rag_model.rag.generator.config.prefix , n_docs=rag_model.config.n_docs , return_tensors="pt" , )
a__ : List[Any] = rag_model.retriever.index.get_doc_dicts(result.doc_ids )
a__ : int = []
for docs in all_docs:
a__ : Optional[int] = [strip_title(__a ) for title in docs["title"]]
provenance_strings.append("\t".join(__a ) )
return provenance_strings
def UpperCamelCase_ ( __a , __a , __a ) -> Dict:
with torch.no_grad():
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a )
a__ : Any = inputs_dict.input_ids.to(args.device )
a__ : Dict = inputs_dict.attention_mask.to(args.device )
a__ : Optional[int] = rag_model.generate( # rag_model overwrites generate
__a , attention_mask=__a , num_beams=args.num_beams , min_length=args.min_length , max_length=args.max_length , early_stopping=__a , num_return_sequences=1 , bad_words_ids=[[0, 0]] , )
a__ : int = rag_model.retriever.generator_tokenizer.batch_decode(__a , skip_special_tokens=__a )
if args.print_predictions:
for q, a in zip(__a , __a ):
logger.info("Q: {} - A: {}".format(__a , __a ) )
return answers
def UpperCamelCase_ ( ) -> List[str]:
a__ : int = argparse.ArgumentParser()
parser.add_argument(
"--model_type" , choices=["rag_sequence", "rag_token", "bart"] , type=__a , help=(
"RAG model type: rag_sequence, rag_token or bart, if none specified, the type is inferred from the"
" model_name_or_path"
) , )
parser.add_argument(
"--index_name" , default=__a , choices=["exact", "compressed", "legacy"] , type=__a , help="RAG model retriever type" , )
parser.add_argument(
"--index_path" , default=__a , type=__a , help="Path to the retrieval index" , )
parser.add_argument("--n_docs" , default=5 , type=__a , help="Number of retrieved docs" )
parser.add_argument(
"--model_name_or_path" , default=__a , type=__a , required=__a , help="Path to pretrained checkpoints or model identifier from huggingface.co/models" , )
parser.add_argument(
"--eval_mode" , choices=["e2e", "retrieval"] , default="e2e" , type=__a , help=(
"Evaluation mode, e2e calculates exact match and F1 of the downstream task, retrieval calculates"
" precision@k."
) , )
parser.add_argument("--k" , default=1 , type=__a , help="k for the precision@k calculation" )
parser.add_argument(
"--evaluation_set" , default=__a , type=__a , required=__a , help="Path to a file containing evaluation samples" , )
parser.add_argument(
"--gold_data_path" , default=__a , type=__a , required=__a , help="Path to a tab-separated file with gold samples" , )
parser.add_argument(
"--gold_data_mode" , default="qa" , type=__a , choices=["qa", "ans"] , help=(
"Format of the gold data file"
"qa - a single line in the following format: question [tab] answer_list"
"ans - a single line of the gold file contains the expected answer string"
) , )
parser.add_argument(
"--predictions_path" , type=__a , default="predictions.txt" , help="Name of the predictions file, to be stored in the checkpoints directory" , )
parser.add_argument(
"--eval_all_checkpoints" , action="store_true" , help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number" , )
parser.add_argument(
"--eval_batch_size" , default=8 , type=__a , help="Batch size per GPU/CPU for evaluation." , )
parser.add_argument(
"--recalculate" , help="Recalculate predictions even if the prediction file exists" , action="store_true" , )
parser.add_argument(
"--num_beams" , default=4 , type=__a , help="Number of beams to be used when generating answers" , )
parser.add_argument("--min_length" , default=1 , type=__a , help="Min length of the generated answers" )
parser.add_argument("--max_length" , default=50 , type=__a , help="Max length of the generated answers" )
parser.add_argument(
"--print_predictions" , action="store_true" , help="If True, prints predictions while evaluating." , )
parser.add_argument(
"--print_docs" , action="store_true" , help="If True, prints docs retried while generating." , )
a__ : int = parser.parse_args()
a__ : Dict = torch.device("cuda" if torch.cuda.is_available() else "cpu" )
return args
def UpperCamelCase_ ( __a ) -> Optional[int]:
a__ : Tuple = {}
if args.model_type is None:
a__ : List[str] = infer_model_type(args.model_name_or_path )
assert args.model_type is not None
if args.model_type.startswith("rag" ):
a__ : int = RagTokenForGeneration if args.model_type == "rag_token" else RagSequenceForGeneration
a__ : Tuple = args.n_docs
if args.index_name is not None:
a__ : Any = args.index_name
if args.index_path is not None:
a__ : int = args.index_path
else:
a__ : Optional[Any] = BartForConditionalGeneration
a__ : Tuple = (
[f.path for f in os.scandir(args.model_name_or_path ) if f.is_dir()]
if args.eval_all_checkpoints
else [args.model_name_or_path]
)
logger.info("Evaluate the following checkpoints: %s" , __a )
a__ : Any = get_scores if args.eval_mode == "e2e" else get_precision_at_k
a__ : Union[str, Any] = evaluate_batch_eae if args.eval_mode == "e2e" else evaluate_batch_retrieval
for checkpoint in checkpoints:
if os.path.exists(args.predictions_path ) and (not args.recalculate):
logger.info("Calculating metrics based on an existing predictions file: {}".format(args.predictions_path ) )
score_fn(__a , args.predictions_path , args.gold_data_path )
continue
logger.info("***** Running evaluation for {} *****".format(__a ) )
logger.info(" Batch size = %d" , args.eval_batch_size )
logger.info(" Predictions will be stored under {}".format(args.predictions_path ) )
if args.model_type.startswith("rag" ):
a__ : str = RagRetriever.from_pretrained(__a , **__a )
a__ : Optional[int] = model_class.from_pretrained(__a , retriever=__a , **__a )
model.retriever.init_retrieval()
else:
a__ : Dict = model_class.from_pretrained(__a , **__a )
model.to(args.device )
with open(args.evaluation_set , "r" ) as eval_file, open(args.predictions_path , "w" ) as preds_file:
a__ : List[Any] = []
for line in tqdm(__a ):
questions.append(line.strip() )
if len(__a ) == args.eval_batch_size:
a__ : Union[str, Any] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) + "\n" )
preds_file.flush()
a__ : Any = []
if len(__a ) > 0:
a__ : List[str] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) )
preds_file.flush()
score_fn(__a , args.predictions_path , args.gold_data_path )
if __name__ == "__main__":
UpperCamelCase : List[Any] = get_args()
main(args)
| 37 | 1 |
from typing import Mapping
from ...configuration_utils import PretrainedConfig
from ...onnx import OnnxSeqaSeqConfigWithPast
from ...utils import logging
UpperCamelCase : Tuple = logging.get_logger(__name__)
UpperCamelCase : List[Any] = {
"""google/umt5-small""": """https://huggingface.co/google/umt5-small/resolve/main/config.json""",
# See all umt5 models at https://huggingface.co/models?filter=umt5
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'umt5'
_lowercase = ['past_key_values']
def __init__( self : Union[str, Any] , lowerCamelCase__ : List[Any]=250_112 , lowerCamelCase__ : str=512 , lowerCamelCase__ : Optional[Any]=64 , lowerCamelCase__ : List[Any]=1_024 , lowerCamelCase__ : int=8 , lowerCamelCase__ : List[str]=None , lowerCamelCase__ : Optional[Any]=6 , lowerCamelCase__ : int=32 , lowerCamelCase__ : Union[str, Any]=128 , lowerCamelCase__ : Tuple=0.1 , lowerCamelCase__ : Optional[Any]=1E-6 , lowerCamelCase__ : Union[str, Any]=1.0 , lowerCamelCase__ : Union[str, Any]="gated-gelu" , lowerCamelCase__ : Any=True , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : Any="T5Tokenizer" , lowerCamelCase__ : Union[str, Any]=True , lowerCamelCase__ : Optional[Any]=0 , lowerCamelCase__ : List[Any]=1 , lowerCamelCase__ : List[str]=0 , **lowerCamelCase__ : Any , ):
super().__init__(
is_encoder_decoder=lowerCamelCase__ , tokenizer_class=lowerCamelCase__ , tie_word_embeddings=lowerCamelCase__ , pad_token_id=lowerCamelCase__ , eos_token_id=lowerCamelCase__ , decoder_start_token_id=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : Dict = vocab_size
a__ : List[Any] = d_model
a__ : Optional[Any] = d_kv
a__ : Optional[int] = d_ff
a__ : List[Any] = num_layers
a__ : int = (
num_decoder_layers if num_decoder_layers is not None else self.num_layers
) # default = symmetry
a__ : str = num_heads
a__ : Optional[int] = relative_attention_num_buckets
a__ : Any = relative_attention_max_distance
a__ : int = dropout_rate
a__ : Union[str, Any] = layer_norm_epsilon
a__ : Optional[int] = initializer_factor
a__ : Optional[int] = feed_forward_proj
a__ : Tuple = use_cache
a__ : List[Any] = self.feed_forward_proj.split("-" )
a__ : Optional[int] = act_info[-1]
a__ : Union[str, Any] = act_info[0] == "gated"
if len(lowerCamelCase__ ) > 1 and act_info[0] != "gated" or len(lowerCamelCase__ ) > 2:
raise ValueError(
f'''`feed_forward_proj`: {feed_forward_proj} is not a valid activation function of the dense layer.'''
"Please make sure `feed_forward_proj` is of the format `gated-{ACT_FN}` or `{ACT_FN}`, e.g. "
"'gated-gelu' or 'relu'" )
if feed_forward_proj == "gated-gelu":
a__ : Optional[int] = "gelu_new"
@property
def _UpperCamelCase( self : int ):
return self.d_model
@property
def _UpperCamelCase( self : int ):
return self.num_heads
@property
def _UpperCamelCase( self : Dict ):
return self.num_layers
class A__ ( A__ ):
"""simple docstring"""
@property
# Copied from transformers.models.t5.configuration_t5.T5OnnxConfig.inputs
def _UpperCamelCase( self : Optional[Any] ):
a__ : Dict = {
"input_ids": {0: "batch", 1: "encoder_sequence"},
"attention_mask": {0: "batch", 1: "encoder_sequence"},
}
if self.use_past:
a__ : Tuple = "past_encoder_sequence + sequence"
a__ : Dict = {0: "batch"}
a__ : Union[str, Any] = {0: "batch", 1: "past_decoder_sequence + sequence"}
else:
a__ : Tuple = {0: "batch", 1: "decoder_sequence"}
a__ : List[str] = {0: "batch", 1: "decoder_sequence"}
if self.use_past:
self.fill_with_past_key_values_(lowerCamelCase__ , direction="inputs" )
return common_inputs
@property
# Copied from transformers.models.t5.configuration_t5.T5OnnxConfig.default_onnx_opset
def _UpperCamelCase( self : int ):
return 13
@property
def _UpperCamelCase( self : List[str] ):
return 5E-4
| 37 |
import logging
import os
from dataclasses import dataclass, field
from typing import Dict, Optional
import datasets
import numpy as np
import tensorflow as tf
from transformers import (
AutoConfig,
AutoTokenizer,
EvalPrediction,
HfArgumentParser,
PreTrainedTokenizer,
TFAutoModelForSequenceClassification,
TFTrainer,
TFTrainingArguments,
)
from transformers.utils import logging as hf_logging
hf_logging.set_verbosity_info()
hf_logging.enable_default_handler()
hf_logging.enable_explicit_format()
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a = None , ) -> str:
a__ : int = {}
if train_file is not None:
a__ : int = [train_file]
if eval_file is not None:
a__ : Union[str, Any] = [eval_file]
if test_file is not None:
a__ : str = [test_file]
a__ : Optional[Any] = datasets.load_dataset("csv" , data_files=__a )
a__ : List[Any] = list(ds[list(files.keys() )[0]].features.keys() )
a__ : str = features_name.pop(__a )
a__ : Dict = list(set(ds[list(files.keys() )[0]][label_name] ) )
a__ : str = {label: i for i, label in enumerate(__a )}
a__ : Tuple = tokenizer.model_input_names
a__ : List[str] = {}
if len(__a ) == 1:
for k in files.keys():
a__ : Optional[Any] = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
example[features_name[0]] , truncation=__a , max_length=__a , padding="max_length" ) , batched=__a , )
elif len(__a ) == 2:
for k in files.keys():
a__ : Dict = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
(example[features_name[0]], example[features_name[1]]) , truncation=__a , max_length=__a , padding="max_length" , ) , batched=__a , )
def gen_train():
for ex in transformed_ds[datasets.Split.TRAIN]:
a__ : str = {k: v for k, v in ex.items() if k in input_names}
a__ : str = labelaid[ex[label_name]]
yield (d, label)
def gen_val():
for ex in transformed_ds[datasets.Split.VALIDATION]:
a__ : Tuple = {k: v for k, v in ex.items() if k in input_names}
a__ : List[Any] = labelaid[ex[label_name]]
yield (d, label)
def gen_test():
for ex in transformed_ds[datasets.Split.TEST]:
a__ : List[Any] = {k: v for k, v in ex.items() if k in input_names}
a__ : Optional[int] = labelaid[ex[label_name]]
yield (d, label)
a__ : Optional[Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TRAIN in transformed_ds
else None
)
if train_ds is not None:
a__ : Optional[int] = train_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TRAIN] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.VALIDATION in transformed_ds
else None
)
if val_ds is not None:
a__ : Optional[Any] = val_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.VALIDATION] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TEST in transformed_ds
else None
)
if test_ds is not None:
a__ : Tuple = test_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TEST] ) ) )
return train_ds, val_ds, test_ds, labelaid
UpperCamelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(metadata={'help': 'Which column contains the label'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the training file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the development file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the test file'} )
_lowercase = field(
default=1_2_8 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
_lowercase = field(
default=A__ , metadata={'help': 'Overwrite the cached training and evaluation sets'} )
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
_lowercase = field(default=A__ , metadata={'help': 'Set this flag to use fast tokenization.'} )
# If you want to tweak more attributes on your tokenizer, you should do it in a distinct script,
# or just modify its tokenizer_config.json.
_lowercase = field(
default=A__ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
def UpperCamelCase_ ( ) -> Union[str, Any]:
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
a__ : str = HfArgumentParser((ModelArguments, DataTrainingArguments, TFTrainingArguments) )
a__, a__, a__ : str = parser.parse_args_into_dataclasses()
if (
os.path.exists(training_args.output_dir )
and os.listdir(training_args.output_dir )
and training_args.do_train
and not training_args.overwrite_output_dir
):
raise ValueError(
f'''Output directory ({training_args.output_dir}) already exists and is not empty. Use'''
" --overwrite_output_dir to overcome." )
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s" , datefmt="%m/%d/%Y %H:%M:%S" , level=logging.INFO , )
logger.info(
f'''n_replicas: {training_args.n_replicas}, distributed training: {bool(training_args.n_replicas > 1 )}, '''
f'''16-bits training: {training_args.fpaa}''' )
logger.info(f'''Training/evaluation parameters {training_args}''' )
# Load pretrained model and tokenizer
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
a__ : Union[str, Any] = AutoTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , )
a__, a__, a__, a__ : Optional[Any] = get_tfds(
train_file=data_args.train_file , eval_file=data_args.dev_file , test_file=data_args.test_file , tokenizer=__a , label_column_id=data_args.label_column_id , max_seq_length=data_args.max_seq_length , )
a__ : Optional[int] = AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=len(__a ) , labelaid=__a , idalabel={id: label for label, id in labelaid.items()} , finetuning_task="text-classification" , cache_dir=model_args.cache_dir , )
with training_args.strategy.scope():
a__ : Any = TFAutoModelForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_pt=bool(".bin" in model_args.model_name_or_path ) , config=__a , cache_dir=model_args.cache_dir , )
def compute_metrics(__a ) -> Dict:
a__ : Union[str, Any] = np.argmax(p.predictions , axis=1 )
return {"acc": (preds == p.label_ids).mean()}
# Initialize our Trainer
a__ : Dict = TFTrainer(
model=__a , args=__a , train_dataset=__a , eval_dataset=__a , compute_metrics=__a , )
# Training
if training_args.do_train:
trainer.train()
trainer.save_model()
tokenizer.save_pretrained(training_args.output_dir )
# Evaluation
a__ : Optional[Any] = {}
if training_args.do_eval:
logger.info("*** Evaluate ***" )
a__ : Dict = trainer.evaluate()
a__ : int = os.path.join(training_args.output_dir , "eval_results.txt" )
with open(__a , "w" ) as writer:
logger.info("***** Eval results *****" )
for key, value in result.items():
logger.info(f''' {key} = {value}''' )
writer.write(f'''{key} = {value}\n''' )
results.update(__a )
return results
if __name__ == "__main__":
main()
| 37 | 1 |
from ..utils import DummyObject, requires_backends
class A__ ( metaclass=A__ ):
"""simple docstring"""
_lowercase = ['transformers', 'torch', 'note_seq']
def __init__( self : Optional[int] , *lowerCamelCase__ : int , **lowerCamelCase__ : int ):
requires_backends(self , ["transformers", "torch", "note_seq"] )
@classmethod
def _UpperCamelCase( cls : str , *lowerCamelCase__ : Union[str, Any] , **lowerCamelCase__ : Optional[int] ):
requires_backends(cls , ["transformers", "torch", "note_seq"] )
@classmethod
def _UpperCamelCase( cls : str , *lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Any ):
requires_backends(cls , ["transformers", "torch", "note_seq"] )
| 37 |
import argparse
import collections
import json
import os
import re
import string
import sys
import numpy as np
UpperCamelCase : List[str] = re.compile(r"""\b(a|an|the)\b""", re.UNICODE)
UpperCamelCase : Union[str, Any] = None
def UpperCamelCase_ ( ) -> List[str]:
a__ : List[Any] = argparse.ArgumentParser("Official evaluation script for SQuAD version 2.0." )
parser.add_argument("data_file" , metavar="data.json" , help="Input data JSON file." )
parser.add_argument("pred_file" , metavar="pred.json" , help="Model predictions." )
parser.add_argument(
"--out-file" , "-o" , metavar="eval.json" , help="Write accuracy metrics to file (default is stdout)." )
parser.add_argument(
"--na-prob-file" , "-n" , metavar="na_prob.json" , help="Model estimates of probability of no answer." )
parser.add_argument(
"--na-prob-thresh" , "-t" , type=__a , default=1.0 , help="Predict \"\" if no-answer probability exceeds this (default = 1.0)." , )
parser.add_argument(
"--out-image-dir" , "-p" , metavar="out_images" , default=__a , help="Save precision-recall curves to directory." )
parser.add_argument("--verbose" , "-v" , action="store_true" )
if len(sys.argv ) == 1:
parser.print_help()
sys.exit(1 )
return parser.parse_args()
def UpperCamelCase_ ( __a ) -> str:
a__ : Optional[Any] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : Dict = bool(qa["answers"]["text"] )
return qid_to_has_ans
def UpperCamelCase_ ( __a ) -> List[Any]:
def remove_articles(__a ):
return ARTICLES_REGEX.sub(" " , __a )
def white_space_fix(__a ):
return " ".join(text.split() )
def remove_punc(__a ):
a__ : Union[str, Any] = set(string.punctuation )
return "".join(ch for ch in text if ch not in exclude )
def lower(__a ):
return text.lower()
return white_space_fix(remove_articles(remove_punc(lower(__a ) ) ) )
def UpperCamelCase_ ( __a ) -> Dict:
if not s:
return []
return normalize_answer(__a ).split()
def UpperCamelCase_ ( __a , __a ) -> str:
return int(normalize_answer(__a ) == normalize_answer(__a ) )
def UpperCamelCase_ ( __a , __a ) -> Dict:
a__ : int = get_tokens(__a )
a__ : Optional[Any] = get_tokens(__a )
a__ : Any = collections.Counter(__a ) & collections.Counter(__a )
a__ : Dict = sum(common.values() )
if len(__a ) == 0 or len(__a ) == 0:
# If either is no-answer, then F1 is 1 if they agree, 0 otherwise
return int(gold_toks == pred_toks )
if num_same == 0:
return 0
a__ : Tuple = 1.0 * num_same / len(__a )
a__ : str = 1.0 * num_same / len(__a )
a__ : str = (2 * precision * recall) / (precision + recall)
return fa
def UpperCamelCase_ ( __a , __a ) -> int:
a__ : List[str] = {}
a__ : Optional[int] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : List[Any] = qa["id"]
a__ : Dict = [t for t in qa["answers"]["text"] if normalize_answer(__a )]
if not gold_answers:
# For unanswerable questions, only correct answer is empty string
a__ : Tuple = [""]
if qid not in preds:
print(f'''Missing prediction for {qid}''' )
continue
a__ : Tuple = preds[qid]
# Take max over all gold answers
a__ : Optional[int] = max(compute_exact(__a , __a ) for a in gold_answers )
a__ : str = max(compute_fa(__a , __a ) for a in gold_answers )
return exact_scores, fa_scores
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
a__ : Optional[Any] = {}
for qid, s in scores.items():
a__ : Dict = na_probs[qid] > na_prob_thresh
if pred_na:
a__ : Dict = float(not qid_to_has_ans[qid] )
else:
a__ : Optional[Any] = s
return new_scores
def UpperCamelCase_ ( __a , __a , __a=None ) -> Tuple:
if not qid_list:
a__ : Union[str, Any] = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores.values() ) / total),
("f1", 100.0 * sum(fa_scores.values() ) / total),
("total", total),
] )
else:
a__ : int = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores[k] for k in qid_list ) / total),
("f1", 100.0 * sum(fa_scores[k] for k in qid_list ) / total),
("total", total),
] )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
for k in new_eval:
a__ : Optional[Any] = new_eval[k]
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
plt.step(__a , __a , color="b" , alpha=0.2 , where="post" )
plt.fill_between(__a , __a , step="post" , alpha=0.2 , color="b" )
plt.xlabel("Recall" )
plt.ylabel("Precision" )
plt.xlim([0.0, 1.05] )
plt.ylim([0.0, 1.05] )
plt.title(__a )
plt.savefig(__a )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a , __a=None , __a=None ) -> Dict:
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
a__ : Any = 0.0
a__ : Optional[int] = 1.0
a__ : Optional[int] = 0.0
a__ : Any = [1.0]
a__ : Tuple = [0.0]
a__ : List[str] = 0.0
for i, qid in enumerate(__a ):
if qid_to_has_ans[qid]:
true_pos += scores[qid]
a__ : Any = true_pos / float(i + 1 )
a__ : int = true_pos / float(__a )
if i == len(__a ) - 1 or na_probs[qid] != na_probs[qid_list[i + 1]]:
# i.e., if we can put a threshold after this point
avg_prec += cur_p * (cur_r - recalls[-1])
precisions.append(__a )
recalls.append(__a )
if out_image:
plot_pr_curve(__a , __a , __a , __a )
return {"ap": 100.0 * avg_prec}
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> str:
if out_image_dir and not os.path.exists(__a ):
os.makedirs(__a )
a__ : Optional[int] = sum(1 for v in qid_to_has_ans.values() if v )
if num_true_pos == 0:
return
a__ : Optional[int] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_exact.png" ) , title="Precision-Recall curve for Exact Match score" , )
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_f1.png" ) , title="Precision-Recall curve for F1 score" , )
a__ : str = {k: float(__a ) for k, v in qid_to_has_ans.items()}
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_oracle.png" ) , title="Oracle Precision-Recall curve (binary task of HasAns vs. NoAns)" , )
merge_eval(__a , __a , "pr_exact" )
merge_eval(__a , __a , "pr_f1" )
merge_eval(__a , __a , "pr_oracle" )
def UpperCamelCase_ ( __a , __a , __a , __a ) -> str:
if not qid_list:
return
a__ : Optional[Any] = [na_probs[k] for k in qid_list]
a__ : str = np.ones_like(__a ) / float(len(__a ) )
plt.hist(__a , weights=__a , bins=20 , range=(0.0, 1.0) )
plt.xlabel("Model probability of no-answer" )
plt.ylabel("Proportion of dataset" )
plt.title(f'''Histogram of no-answer probability: {name}''' )
plt.savefig(os.path.join(__a , f'''na_prob_hist_{name}.png''' ) )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[Any]:
a__ : str = sum(1 for k in qid_to_has_ans if not qid_to_has_ans[k] )
a__ : Optional[Any] = num_no_ans
a__ : Dict = cur_score
a__ : Any = 0.0
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
for i, qid in enumerate(__a ):
if qid not in scores:
continue
if qid_to_has_ans[qid]:
a__ : Optional[int] = scores[qid]
else:
if preds[qid]:
a__ : str = -1
else:
a__ : Union[str, Any] = 0
cur_score += diff
if cur_score > best_score:
a__ : Any = cur_score
a__ : Dict = na_probs[qid]
return 100.0 * best_score / len(__a ), best_thresh
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> Any:
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__ : Any = best_exact
a__ : Any = exact_thresh
a__ : List[Any] = best_fa
a__ : Optional[int] = fa_thresh
def UpperCamelCase_ ( ) -> Tuple:
with open(OPTS.data_file ) as f:
a__ : List[Any] = json.load(__a )
a__ : Any = dataset_json["data"]
with open(OPTS.pred_file ) as f:
a__ : int = json.load(__a )
if OPTS.na_prob_file:
with open(OPTS.na_prob_file ) as f:
a__ : List[str] = json.load(__a )
else:
a__ : Optional[int] = {k: 0.0 for k in preds}
a__ : Optional[Any] = make_qid_to_has_ans(__a ) # maps qid to True/False
a__ : List[Any] = [k for k, v in qid_to_has_ans.items() if v]
a__ : Union[str, Any] = [k for k, v in qid_to_has_ans.items() if not v]
a__, a__ : str = get_raw_scores(__a , __a )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : Tuple = make_eval_dict(__a , __a )
if has_ans_qids:
a__ : str = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "HasAns" )
if no_ans_qids:
a__ : List[Any] = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "NoAns" )
if OPTS.na_prob_file:
find_all_best_thresh(__a , __a , __a , __a , __a , __a )
if OPTS.na_prob_file and OPTS.out_image_dir:
run_precision_recall_analysis(__a , __a , __a , __a , __a , OPTS.out_image_dir )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "hasAns" )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "noAns" )
if OPTS.out_file:
with open(OPTS.out_file , "w" ) as f:
json.dump(__a , __a )
else:
print(json.dumps(__a , indent=2 ) )
if __name__ == "__main__":
UpperCamelCase : Any = parse_args()
if OPTS.out_image_dir:
import matplotlib
matplotlib.use("""Agg""")
import matplotlib.pyplot as plt
main()
| 37 | 1 |
def UpperCamelCase_ ( __a ) -> bool:
a__ : List[Any] = 0
for ch in input_str:
a__ : str = ord(__a )
a__ : Any = pow(2 , __a )
# If we already turned on bit for current character's unicode
if bitmap >> ch_unicode & 1 == 1:
return False
bitmap |= ch_bit_index_on
return True
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 |
import json
import os
import unittest
from transformers import CLIPTokenizer, CLIPTokenizerFast
from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_ftfy, require_tokenizers
from ...test_tokenization_common import TokenizerTesterMixin
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = CLIPTokenizer
_lowercase = CLIPTokenizerFast
_lowercase = True
_lowercase = {}
_lowercase = False
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# fmt: off
a__ : Any = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "lo", "l</w>", "w</w>", "r</w>", "t</w>", "low</w>", "er</w>", "lowest</w>", "newer</w>", "wider", "<unk>", "<|startoftext|>", "<|endoftext|>"]
# fmt: on
a__ : Optional[Any] = dict(zip(lowerCamelCase__ , range(len(lowerCamelCase__ ) ) ) )
a__ : Optional[Any] = ["#version: 0.2", "l o", "lo w</w>", "e r</w>"]
a__ : Optional[Any] = {"unk_token": "<unk>"}
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["vocab_file"] )
a__ : int = os.path.join(self.tmpdirname , VOCAB_FILES_NAMES["merges_file"] )
with open(self.vocab_file , "w" , encoding="utf-8" ) as fp:
fp.write(json.dumps(lowerCamelCase__ ) + "\n" )
with open(self.merges_file , "w" , encoding="utf-8" ) as fp:
fp.write("\n".join(lowerCamelCase__ ) )
def _UpperCamelCase( self : Dict , **lowerCamelCase__ : int ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizer.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] , **lowerCamelCase__ : Optional[int] ):
kwargs.update(self.special_tokens_map )
return CLIPTokenizerFast.from_pretrained(self.tmpdirname , **lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[Any] ):
a__ : int = "lower newer"
a__ : Optional[int] = "lower newer"
return input_text, output_text
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = CLIPTokenizer(self.vocab_file , self.merges_file , **self.special_tokens_map )
a__ : int = "lower newer"
a__ : List[str] = ["lo", "w", "er</w>", "n", "e", "w", "er</w>"]
a__ : Union[str, Any] = tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : int = tokens + [tokenizer.unk_token]
a__ : Union[str, Any] = [10, 2, 16, 9, 3, 2, 16, 20]
self.assertListEqual(tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , lowerCamelCase__ )
@require_ftfy
def _UpperCamelCase( self : Optional[Any] ):
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : List[str] = self.tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : Any = self.rust_tokenizer_class.from_pretrained(lowerCamelCase__ , **lowerCamelCase__ )
a__ : int = "A\n'll 11p223RF☆ho!!to?'d'd''d of a cat to-$''d."
a__ : Optional[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : Dict = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on an example containing a character (Latin Small Letter A
# with Tilde) encoded in 2 different ways
a__ : Optional[Any] = "xa\u0303y" + " " + "x\xe3y"
a__ : Optional[int] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of space type
a__ : str = [
"\u0009", # (horizontal tab, '\t')
"\u000B", # (vertical tab)
"\u000C", # (form feed)
"\u0020", # (space, ' ')
"\u200E", # (left-to-right mark):w
"\u200F", # (right-to-left mark)
]
for unicode_seq in spaces_unicodes:
a__ : Any = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
# Test that the tokenization is identical on unicode of line break type
a__ : Union[str, Any] = [
"\u000A", # (line feed, '\n')
"\r\n", # (carriage return and line feed, '\r\n')
"\u000D", # (carriage return, '\r')
"\r", # (carriage return, '\r')
"\u000D", # (carriage return, '\r')
"\u2028", # (line separator)
"\u2029", # (paragraph separator)
# "\u0085", # (next line)
]
# The tokenization is not identical for the character "\u0085" (next line). The slow version using ftfy transforms
# it into the Horizontal Ellipsis character "…" ("\u2026") while the fast version transforms it into a
# space (and thus into an empty list).
for unicode_seq in line_break_unicodes:
a__ : List[Any] = tokenizer_s.tokenize(lowerCamelCase__ )
a__ : int = tokenizer_r.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
# Test which aims to verify that the offsets are well adapted to the argument `add_prefix_space`
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f'''{tokenizer.__class__.__name__} ({pretrained_name})''' ):
a__ : str = "hello" # `hello` is a token in the vocabulary of `pretrained_name`
a__ : Tuple = f'''{text_of_1_token} {text_of_1_token}'''
a__ : Optional[int] = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Union[str, Any] = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (0, len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (len(lowerCamelCase__ ) + 1, len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
a__ : Optional[Any] = f''' {text}'''
a__ : str = self.rust_tokenizer_class.from_pretrained(
lowerCamelCase__ , use_fast=lowerCamelCase__ , )
a__ : Dict = tokenizer_r(lowerCamelCase__ , return_offsets_mapping=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertEqual(encoding.offset_mapping[0] , (1, 1 + len(lowerCamelCase__ )) )
self.assertEqual(
encoding.offset_mapping[1] , (1 + len(lowerCamelCase__ ) + 1, 1 + len(lowerCamelCase__ ) + 1 + len(lowerCamelCase__ )) , )
def _UpperCamelCase( self : int ):
# Test related to the breaking change introduced in transformers v4.17.0
# We need to check that an error in raised when the user try to load a previous version of the tokenizer.
with self.assertRaises(lowerCamelCase__ ) as context:
self.rust_tokenizer_class.from_pretrained("robot-test/old-clip-tokenizer" )
self.assertTrue(
context.exception.args[0].startswith(
"The `backend_tokenizer` provided does not match the expected format." ) )
@require_ftfy
def _UpperCamelCase( self : int ):
super().test_tokenization_python_rust_equals()
def _UpperCamelCase( self : str ):
# CLIP always lower cases letters
pass
| 37 | 1 |
from typing import List, Optional, Union
import torch
from transformers import (
XLMRobertaTokenizer,
)
from ...models import UNetaDConditionModel, VQModel
from ...pipelines import DiffusionPipeline
from ...pipelines.pipeline_utils import ImagePipelineOutput
from ...schedulers import DDIMScheduler, DDPMScheduler
from ...utils import (
is_accelerate_available,
is_accelerate_version,
logging,
randn_tensor,
replace_example_docstring,
)
from .text_encoder import MultilingualCLIP
UpperCamelCase : Tuple = logging.get_logger(__name__) # pylint: disable=invalid-name
UpperCamelCase : str = """
Examples:
```py
>>> from diffusers import KandinskyPipeline, KandinskyPriorPipeline
>>> import torch
>>> pipe_prior = KandinskyPriorPipeline.from_pretrained(\"kandinsky-community/Kandinsky-2-1-prior\")
>>> pipe_prior.to(\"cuda\")
>>> prompt = \"red cat, 4k photo\"
>>> out = pipe_prior(prompt)
>>> image_emb = out.image_embeds
>>> negative_image_emb = out.negative_image_embeds
>>> pipe = KandinskyPipeline.from_pretrained(\"kandinsky-community/kandinsky-2-1\")
>>> pipe.to(\"cuda\")
>>> image = pipe(
... prompt,
... image_embeds=image_emb,
... negative_image_embeds=negative_image_emb,
... height=768,
... width=768,
... num_inference_steps=100,
... ).images
>>> image[0].save(\"cat.png\")
```
"""
def UpperCamelCase_ ( __a , __a , __a=8 ) -> List[str]:
a__ : List[str] = h // scale_factor**2
if h % scale_factor**2 != 0:
new_h += 1
a__ : Optional[Any] = w // scale_factor**2
if w % scale_factor**2 != 0:
new_w += 1
return new_h * scale_factor, new_w * scale_factor
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Dict , lowerCamelCase__ : MultilingualCLIP , lowerCamelCase__ : XLMRobertaTokenizer , lowerCamelCase__ : UNetaDConditionModel , lowerCamelCase__ : Union[DDIMScheduler, DDPMScheduler] , lowerCamelCase__ : VQModel , ):
super().__init__()
self.register_modules(
text_encoder=lowerCamelCase__ , tokenizer=lowerCamelCase__ , unet=lowerCamelCase__ , scheduler=lowerCamelCase__ , movq=lowerCamelCase__ , )
a__ : int = 2 ** (len(self.movq.config.block_out_channels ) - 1)
def _UpperCamelCase( self : Any , lowerCamelCase__ : Tuple , lowerCamelCase__ : Any , lowerCamelCase__ : Tuple , lowerCamelCase__ : int , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[int] ):
if latents is None:
a__ : Tuple = randn_tensor(lowerCamelCase__ , generator=lowerCamelCase__ , device=lowerCamelCase__ , dtype=lowerCamelCase__ )
else:
if latents.shape != shape:
raise ValueError(f'''Unexpected latents shape, got {latents.shape}, expected {shape}''' )
a__ : Optional[Any] = latents.to(lowerCamelCase__ )
a__ : Tuple = latents * scheduler.init_noise_sigma
return latents
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Any , lowerCamelCase__ : List[Any] , lowerCamelCase__ : List[str] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[str]=None , ):
a__ : Union[str, Any] = len(lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else 1
# get prompt text embeddings
a__ : List[Any] = self.tokenizer(
lowerCamelCase__ , padding="max_length" , truncation=lowerCamelCase__ , max_length=77 , return_attention_mask=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ , return_tensors="pt" , )
a__ : Tuple = text_inputs.input_ids
a__ : Union[str, Any] = self.tokenizer(lowerCamelCase__ , padding="longest" , return_tensors="pt" ).input_ids
if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(lowerCamelCase__ , lowerCamelCase__ ):
a__ : Optional[int] = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1] )
logger.warning(
"The following part of your input was truncated because CLIP can only handle sequences up to"
f''' {self.tokenizer.model_max_length} tokens: {removed_text}''' )
a__ : Optional[Any] = text_input_ids.to(lowerCamelCase__ )
a__ : List[Any] = text_inputs.attention_mask.to(lowerCamelCase__ )
a__, a__ : str = self.text_encoder(
input_ids=lowerCamelCase__ , attention_mask=lowerCamelCase__ )
a__ : Any = prompt_embeds.repeat_interleave(lowerCamelCase__ , dim=0 )
a__ : Optional[Any] = text_encoder_hidden_states.repeat_interleave(lowerCamelCase__ , dim=0 )
a__ : Optional[int] = text_mask.repeat_interleave(lowerCamelCase__ , dim=0 )
if do_classifier_free_guidance:
a__ : List[str]
if negative_prompt is None:
a__ : Optional[int] = [""] * batch_size
elif type(lowerCamelCase__ ) is not type(lowerCamelCase__ ):
raise TypeError(
f'''`negative_prompt` should be the same type to `prompt`, but got {type(lowerCamelCase__ )} !='''
f''' {type(lowerCamelCase__ )}.''' )
elif isinstance(lowerCamelCase__ , lowerCamelCase__ ):
a__ : Dict = [negative_prompt]
elif batch_size != len(lowerCamelCase__ ):
raise ValueError(
f'''`negative_prompt`: {negative_prompt} has batch size {len(lowerCamelCase__ )}, but `prompt`:'''
f''' {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches'''
" the batch size of `prompt`." )
else:
a__ : Any = negative_prompt
a__ : Any = self.tokenizer(
lowerCamelCase__ , padding="max_length" , max_length=77 , truncation=lowerCamelCase__ , return_attention_mask=lowerCamelCase__ , add_special_tokens=lowerCamelCase__ , return_tensors="pt" , )
a__ : Dict = uncond_input.input_ids.to(lowerCamelCase__ )
a__ : Dict = uncond_input.attention_mask.to(lowerCamelCase__ )
a__, a__ : Any = self.text_encoder(
input_ids=lowerCamelCase__ , attention_mask=lowerCamelCase__ )
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
a__ : Tuple = negative_prompt_embeds.shape[1]
a__ : List[Any] = negative_prompt_embeds.repeat(1 , lowerCamelCase__ )
a__ : Dict = negative_prompt_embeds.view(batch_size * num_images_per_prompt , lowerCamelCase__ )
a__ : List[str] = uncond_text_encoder_hidden_states.shape[1]
a__ : Dict = uncond_text_encoder_hidden_states.repeat(1 , lowerCamelCase__ , 1 )
a__ : List[str] = uncond_text_encoder_hidden_states.view(
batch_size * num_images_per_prompt , lowerCamelCase__ , -1 )
a__ : List[str] = uncond_text_mask.repeat_interleave(lowerCamelCase__ , dim=0 )
# done duplicates
# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch
# to avoid doing two forward passes
a__ : List[str] = torch.cat([negative_prompt_embeds, prompt_embeds] )
a__ : Union[str, Any] = torch.cat([uncond_text_encoder_hidden_states, text_encoder_hidden_states] )
a__ : str = torch.cat([uncond_text_mask, text_mask] )
return prompt_embeds, text_encoder_hidden_states, text_mask
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[str]=0 ):
if is_accelerate_available():
from accelerate import cpu_offload
else:
raise ImportError("Please install accelerate via `pip install accelerate`" )
a__ : str = torch.device(f'''cuda:{gpu_id}''' )
a__ : Dict = [
self.unet,
self.text_encoder,
self.movq,
]
for cpu_offloaded_model in models:
if cpu_offloaded_model is not None:
cpu_offload(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : int , lowerCamelCase__ : str=0 ):
if is_accelerate_available() and is_accelerate_version(">=" , "0.17.0.dev0" ):
from accelerate import cpu_offload_with_hook
else:
raise ImportError("`enable_model_cpu_offload` requires `accelerate v0.17.0` or higher." )
a__ : List[str] = torch.device(f'''cuda:{gpu_id}''' )
if self.device.type != "cpu":
self.to("cpu" , silence_dtype_warnings=lowerCamelCase__ )
torch.cuda.empty_cache() # otherwise we don't see the memory savings (but they probably exist)
a__ : Optional[int] = None
for cpu_offloaded_model in [self.text_encoder, self.unet, self.movq]:
a__, a__ : Optional[Any] = cpu_offload_with_hook(lowerCamelCase__ , lowerCamelCase__ , prev_module_hook=lowerCamelCase__ )
if self.safety_checker is not None:
a__, a__ : List[Any] = cpu_offload_with_hook(self.safety_checker , lowerCamelCase__ , prev_module_hook=lowerCamelCase__ )
# We'll offload the last model manually.
a__ : Optional[int] = hook
@property
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device
def _UpperCamelCase( self : Union[str, Any] ):
if not hasattr(self.unet , "_hf_hook" ):
return self.device
for module in self.unet.modules():
if (
hasattr(lowerCamelCase__ , "_hf_hook" )
and hasattr(module._hf_hook , "execution_device" )
and module._hf_hook.execution_device is not None
):
return torch.device(module._hf_hook.execution_device )
return self.device
@torch.no_grad()
@replace_example_docstring(lowerCamelCase__ )
def __call__( self : List[Any] , lowerCamelCase__ : Union[str, List[str]] , lowerCamelCase__ : Union[torch.FloatTensor, List[torch.FloatTensor]] , lowerCamelCase__ : Union[torch.FloatTensor, List[torch.FloatTensor]] , lowerCamelCase__ : Optional[Union[str, List[str]]] = None , lowerCamelCase__ : int = 512 , lowerCamelCase__ : int = 512 , lowerCamelCase__ : int = 100 , lowerCamelCase__ : float = 4.0 , lowerCamelCase__ : int = 1 , lowerCamelCase__ : Optional[Union[torch.Generator, List[torch.Generator]]] = None , lowerCamelCase__ : Optional[torch.FloatTensor] = None , lowerCamelCase__ : Optional[str] = "pil" , lowerCamelCase__ : bool = True , ):
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
a__ : int = 1
elif isinstance(lowerCamelCase__ , lowerCamelCase__ ):
a__ : List[str] = len(lowerCamelCase__ )
else:
raise ValueError(f'''`prompt` has to be of type `str` or `list` but is {type(lowerCamelCase__ )}''' )
a__ : Dict = self._execution_device
a__ : List[str] = batch_size * num_images_per_prompt
a__ : Tuple = guidance_scale > 1.0
a__, a__, a__ : Optional[int] = self._encode_prompt(
lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
a__ : List[Any] = torch.cat(lowerCamelCase__ , dim=0 )
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
a__ : int = torch.cat(lowerCamelCase__ , dim=0 )
if do_classifier_free_guidance:
a__ : Any = image_embeds.repeat_interleave(lowerCamelCase__ , dim=0 )
a__ : Any = negative_image_embeds.repeat_interleave(lowerCamelCase__ , dim=0 )
a__ : List[str] = torch.cat([negative_image_embeds, image_embeds] , dim=0 ).to(
dtype=prompt_embeds.dtype , device=lowerCamelCase__ )
self.scheduler.set_timesteps(lowerCamelCase__ , device=lowerCamelCase__ )
a__ : Union[str, Any] = self.scheduler.timesteps
a__ : Any = self.unet.config.in_channels
a__, a__ : Optional[int] = get_new_h_w(lowerCamelCase__ , lowerCamelCase__ , self.movq_scale_factor )
# create initial latent
a__ : Tuple = self.prepare_latents(
(batch_size, num_channels_latents, height, width) , text_encoder_hidden_states.dtype , lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , self.scheduler , )
for i, t in enumerate(self.progress_bar(lowerCamelCase__ ) ):
# expand the latents if we are doing classifier free guidance
a__ : Optional[int] = torch.cat([latents] * 2 ) if do_classifier_free_guidance else latents
a__ : Optional[int] = {"text_embeds": prompt_embeds, "image_embeds": image_embeds}
a__ : Dict = self.unet(
sample=lowerCamelCase__ , timestep=lowerCamelCase__ , encoder_hidden_states=lowerCamelCase__ , added_cond_kwargs=lowerCamelCase__ , return_dict=lowerCamelCase__ , )[0]
if do_classifier_free_guidance:
a__, a__ : Union[str, Any] = noise_pred.split(latents.shape[1] , dim=1 )
a__, a__ : Tuple = noise_pred.chunk(2 )
a__, a__ : Dict = variance_pred.chunk(2 )
a__ : Union[str, Any] = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
a__ : List[str] = torch.cat([noise_pred, variance_pred_text] , dim=1 )
if not (
hasattr(self.scheduler.config , "variance_type" )
and self.scheduler.config.variance_type in ["learned", "learned_range"]
):
a__, a__ : Dict = noise_pred.split(latents.shape[1] , dim=1 )
# compute the previous noisy sample x_t -> x_t-1
a__ : Tuple = self.scheduler.step(
lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ , generator=lowerCamelCase__ , ).prev_sample
# post-processing
a__ : Any = self.movq.decode(lowerCamelCase__ , force_not_quantize=lowerCamelCase__ )["sample"]
if output_type not in ["pt", "np", "pil"]:
raise ValueError(f'''Only the output types `pt`, `pil` and `np` are supported not output_type={output_type}''' )
if output_type in ["np", "pil"]:
a__ : Optional[Any] = image * 0.5 + 0.5
a__ : List[str] = image.clamp(0 , 1 )
a__ : Optional[int] = image.cpu().permute(0 , 2 , 3 , 1 ).float().numpy()
if output_type == "pil":
a__ : Tuple = self.numpy_to_pil(lowerCamelCase__ )
if not return_dict:
return (image,)
return ImagePipelineOutput(images=lowerCamelCase__ )
| 37 |
import os
import re
import shutil
from argparse import ArgumentParser, Namespace
from datasets.commands import BaseDatasetsCLICommand
from datasets.utils.logging import get_logger
UpperCamelCase : Dict = """<<<<<<< This should probably be modified because it mentions: """
UpperCamelCase : List[Any] = """=======
>>>>>>>
"""
UpperCamelCase : Optional[Any] = [
"""TextEncoderConfig""",
"""ByteTextEncoder""",
"""SubwordTextEncoder""",
"""encoder_config""",
"""maybe_build_from_corpus""",
"""manual_dir""",
]
UpperCamelCase : Any = [
# (pattern, replacement)
# Order is important here for some replacements
(r"""tfds\.core""", r"""datasets"""),
(r"""tf\.io\.gfile\.GFile""", r"""open"""),
(r"""tf\.([\w\d]+)""", r"""datasets.Value('\1')"""),
(r"""tfds\.features\.Text\(\)""", r"""datasets.Value('string')"""),
(r"""tfds\.features\.Text\(""", r"""datasets.Value('string'),"""),
(r"""features\s*=\s*tfds.features.FeaturesDict\(""", r"""features=datasets.Features("""),
(r"""tfds\.features\.FeaturesDict\(""", r"""dict("""),
(r"""The TensorFlow Datasets Authors""", r"""The TensorFlow Datasets Authors and the HuggingFace Datasets Authors"""),
(r"""tfds\.""", r"""datasets."""),
(r"""dl_manager\.manual_dir""", r"""self.config.data_dir"""),
(r"""self\.builder_config""", r"""self.config"""),
]
def UpperCamelCase_ ( __a ) -> Optional[Any]:
return ConvertCommand(args.tfds_path , args.datasets_directory )
class A__ ( A__ ):
"""simple docstring"""
@staticmethod
def _UpperCamelCase( lowerCamelCase__ : ArgumentParser ):
a__ : List[str] = parser.add_parser(
"convert" , help="Convert a TensorFlow Datasets dataset to a HuggingFace Datasets dataset." , )
train_parser.add_argument(
"--tfds_path" , type=lowerCamelCase__ , required=lowerCamelCase__ , help="Path to a TensorFlow Datasets folder to convert or a single tfds file to convert." , )
train_parser.add_argument(
"--datasets_directory" , type=lowerCamelCase__ , required=lowerCamelCase__ , help="Path to the HuggingFace Datasets folder." )
train_parser.set_defaults(func=lowerCamelCase__ )
def __init__( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str , *lowerCamelCase__ : Tuple ):
a__ : str = get_logger("datasets-cli/converting" )
a__ : Optional[Any] = tfds_path
a__ : Optional[int] = datasets_directory
def _UpperCamelCase( self : int ):
if os.path.isdir(self._tfds_path ):
a__ : List[str] = os.path.abspath(self._tfds_path )
elif os.path.isfile(self._tfds_path ):
a__ : Any = os.path.dirname(self._tfds_path )
else:
raise ValueError("--tfds_path is neither a directory nor a file. Please check path." )
a__ : Dict = os.path.abspath(self._datasets_directory )
self._logger.info(f'''Converting datasets from {abs_tfds_path} to {abs_datasets_path}''' )
a__ : Tuple = []
a__ : str = []
a__ : List[Any] = {}
if os.path.isdir(self._tfds_path ):
a__ : List[str] = os.listdir(lowerCamelCase__ )
else:
a__ : Union[str, Any] = [os.path.basename(self._tfds_path )]
for f_name in file_names:
self._logger.info(f'''Looking at file {f_name}''' )
a__ : Any = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
a__ : Dict = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
if not os.path.isfile(lowerCamelCase__ ) or "__init__" in f_name or "_test" in f_name or ".py" not in f_name:
self._logger.info("Skipping file" )
continue
with open(lowerCamelCase__ , encoding="utf-8" ) as f:
a__ : List[Any] = f.readlines()
a__ : Union[str, Any] = []
a__ : Union[str, Any] = False
a__ : Union[str, Any] = False
a__ : Dict = []
for line in lines:
a__ : Optional[Any] = line
# Convert imports
if "import tensorflow.compat.v2 as tf" in out_line:
continue
elif "@tfds.core" in out_line:
continue
elif "builder=self" in out_line:
continue
elif "import tensorflow_datasets.public_api as tfds" in out_line:
a__ : List[Any] = "import datasets\n"
elif "import tensorflow" in out_line:
# order is important here
a__ : List[str] = ""
continue
elif "from absl import logging" in out_line:
a__ : Dict = "from datasets import logging\n"
elif "getLogger" in out_line:
a__ : List[Any] = out_line.replace("getLogger" , "get_logger" )
elif any(expression in out_line for expression in TO_HIGHLIGHT ):
a__ : List[str] = True
a__ : Dict = list(filter(lambda lowerCamelCase__ : e in out_line , lowerCamelCase__ ) )
out_lines.append(HIGHLIGHT_MESSAGE_PRE + str(lowerCamelCase__ ) + "\n" )
out_lines.append(lowerCamelCase__ )
out_lines.append(lowerCamelCase__ )
continue
else:
for pattern, replacement in TO_CONVERT:
a__ : Tuple = re.sub(lowerCamelCase__ , lowerCamelCase__ , lowerCamelCase__ )
# Take care of saving utilities (to later move them together with main script)
if "tensorflow_datasets" in out_line:
a__ : Optional[int] = re.match(r"from\stensorflow_datasets.*import\s([^\.\r\n]+)" , lowerCamelCase__ )
tfds_imports.extend(imp.strip() for imp in match.group(1 ).split("," ) )
a__ : Optional[Any] = "from . import " + match.group(1 )
# Check we have not forget anything
if "tf." in out_line or "tfds." in out_line or "tensorflow_datasets" in out_line:
raise ValueError(f'''Error converting {out_line.strip()}''' )
if "GeneratorBasedBuilder" in out_line or "BeamBasedBuilder" in out_line:
a__ : Optional[int] = True
out_lines.append(lowerCamelCase__ )
if is_builder or "wmt" in f_name:
# We create a new directory for each dataset
a__ : Dict = f_name.replace(".py" , "" )
a__ : Optional[int] = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
a__ : Any = os.path.join(lowerCamelCase__ , lowerCamelCase__ )
os.makedirs(lowerCamelCase__ , exist_ok=lowerCamelCase__ )
self._logger.info(f'''Adding directory {output_dir}''' )
imports_to_builder_map.update({imp: output_dir for imp in tfds_imports} )
else:
# Utilities will be moved at the end
utils_files.append(lowerCamelCase__ )
if needs_manual_update:
with_manual_update.append(lowerCamelCase__ )
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as f:
f.writelines(lowerCamelCase__ )
self._logger.info(f'''Converted in {output_file}''' )
for utils_file in utils_files:
try:
a__ : Any = os.path.basename(lowerCamelCase__ )
a__ : Optional[int] = imports_to_builder_map[f_name.replace(".py" , "" )]
self._logger.info(f'''Moving {dest_folder} to {utils_file}''' )
shutil.copy(lowerCamelCase__ , lowerCamelCase__ )
except KeyError:
self._logger.error(f'''Cannot find destination folder for {utils_file}. Please copy manually.''' )
if with_manual_update:
for file_path in with_manual_update:
self._logger.warning(
f'''You need to manually update file {file_path} to remove configurations using \'TextEncoderConfig\'.''' )
| 37 | 1 |
import sys
import webbrowser
import requests
from bsa import BeautifulSoup
from fake_useragent import UserAgent
if __name__ == "__main__":
print("""Googling.....""")
UpperCamelCase : Optional[int] = """https://www.google.com/search?q=""" + """ """.join(sys.argv[1:])
UpperCamelCase : Any = requests.get(url, headers={"""UserAgent""": UserAgent().random})
# res.raise_for_status()
with open("""project1a.html""", """wb""") as out_file: # only for knowing the class
for data in res.iter_content(1_0000):
out_file.write(data)
UpperCamelCase : Optional[Any] = BeautifulSoup(res.text, """html.parser""")
UpperCamelCase : Dict = list(soup.select(""".eZt8xd"""))[:5]
print(len(links))
for link in links:
if link.text == "Maps":
webbrowser.open(link.get("""href"""))
else:
webbrowser.open(f"""https://google.com{link.get("href")}""")
| 37 |
import os
from typing import Optional
import fsspec
from fsspec.archive import AbstractArchiveFileSystem
from fsspec.utils import DEFAULT_BLOCK_SIZE
class A__ ( A__ ):
"""simple docstring"""
_lowercase = ''
_lowercase = (
None # protocol passed in prefix to the url. ex: "gzip", for gzip://file.txt::http://foo.bar/file.txt.gz
)
_lowercase = None # compression type in fsspec. ex: "gzip"
_lowercase = None # extension of the filename to strip. ex: "".gz" to get file.txt from file.txt.gz
def __init__( self : List[str] , lowerCamelCase__ : str = "" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , **lowerCamelCase__ : List[str] ):
super().__init__(self , **lowerCamelCase__ )
# always open as "rb" since fsspec can then use the TextIOWrapper to make it work for "r" mode
a__ : str = fsspec.open(
lowerCamelCase__ , mode="rb" , protocol=lowerCamelCase__ , compression=self.compression , client_kwargs={
"requote_redirect_url": False, # see https://github.com/huggingface/datasets/pull/5459
"trust_env": True, # Enable reading proxy env variables.
**(target_options or {}).pop("client_kwargs" , {} ), # To avoid issues if it was already passed.
} , **(target_options or {}) , )
a__ : Optional[int] = os.path.basename(self.file.path.split("::" )[0] )
a__ : int = (
self.compressed_name[: self.compressed_name.rindex("." )]
if "." in self.compressed_name
else self.compressed_name
)
a__ : List[Any] = None
@classmethod
def _UpperCamelCase( cls : int , lowerCamelCase__ : int ):
# compressed file paths are always relative to the archive root
return super()._strip_protocol(lowerCamelCase__ ).lstrip("/" )
def _UpperCamelCase( self : Dict ):
if self.dir_cache is None:
a__ : Dict = {**self.file.fs.info(self.file.path ), "name": self.uncompressed_name}
a__ : int = {f["name"]: f}
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str ):
return self.file.open().read()
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : int=None , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[str]=None , **lowerCamelCase__ : Optional[Any] , ):
a__ : Optional[int] = self._strip_protocol(lowerCamelCase__ )
if mode != "rb":
raise ValueError(f'''Tried to read with mode {mode} on file {self.file.path} opened with mode \'rb\'''' )
return self.file.open()
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'bz2'
_lowercase = 'bz2'
_lowercase = '.bz2'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gzip'
_lowercase = 'gzip'
_lowercase = '.gz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'lz4'
_lowercase = 'lz4'
_lowercase = '.lz4'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'xz'
_lowercase = 'xz'
_lowercase = '.xz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'zstd'
_lowercase = 'zstd'
_lowercase = '.zst'
def __init__( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , lowerCamelCase__ : int = DEFAULT_BLOCK_SIZE , **lowerCamelCase__ : Tuple , ):
super().__init__(
fo=lowerCamelCase__ , mode=lowerCamelCase__ , target_protocol=lowerCamelCase__ , target_options=lowerCamelCase__ , block_size=lowerCamelCase__ , **lowerCamelCase__ , )
# We need to wrap the zstd decompressor to avoid this error in fsspec==2021.7.0 and zstandard==0.15.2:
#
# File "/Users/user/.virtualenvs/hf-datasets/lib/python3.7/site-packages/fsspec/core.py", line 145, in open
# out.close = close
# AttributeError: 'zstd.ZstdDecompressionReader' object attribute 'close' is read-only
#
# see https://github.com/intake/filesystem_spec/issues/725
a__ : Any = self.file.__enter__
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : str ):
a__ : List[Any] = file_
def __enter__( self : str ):
self._file.__enter__()
return self
def __exit__( self : int , *lowerCamelCase__ : List[str] , **lowerCamelCase__ : int ):
self._file.__exit__(*lowerCamelCase__ , **lowerCamelCase__ )
def __iter__( self : List[str] ):
return iter(self._file )
def _UpperCamelCase( self : Any ):
return next(self._file )
def __getattr__( self : Optional[Any] , lowerCamelCase__ : Tuple ):
return getattr(self._file , lowerCamelCase__ )
def fixed_enter(*lowerCamelCase__ : List[str] , **lowerCamelCase__ : str ):
return WrappedFile(_enter(*lowerCamelCase__ , **lowerCamelCase__ ) )
a__ : Any = fixed_enter
| 37 | 1 |
from ...configuration_utils import PretrainedConfig
from ...utils import logging
UpperCamelCase : Optional[int] = logging.get_logger(__name__)
UpperCamelCase : List[str] = {
"""tiiuae/falcon-40b""": """https://huggingface.co/tiiuae/falcon-40b/resolve/main/config.json""",
"""tiiuae/falcon-7b""": """https://huggingface.co/tiiuae/falcon-7b/resolve/main/config.json""",
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'falcon'
_lowercase = ['past_key_values']
def __init__( self : Optional[Any] , lowerCamelCase__ : Union[str, Any]=65_024 , lowerCamelCase__ : List[Any]=4_544 , lowerCamelCase__ : str=32 , lowerCamelCase__ : List[Any]=71 , lowerCamelCase__ : List[Any]=1E-5 , lowerCamelCase__ : List[Any]=0.02 , lowerCamelCase__ : Optional[int]=True , lowerCamelCase__ : Optional[int]=0.0 , lowerCamelCase__ : Optional[int]=0.0 , lowerCamelCase__ : Optional[Any]=None , lowerCamelCase__ : Optional[int]=False , lowerCamelCase__ : Tuple=False , lowerCamelCase__ : List[Any]=True , lowerCamelCase__ : Any=True , lowerCamelCase__ : List[Any]=False , lowerCamelCase__ : int=11 , lowerCamelCase__ : List[Any]=11 , **lowerCamelCase__ : int , ):
a__ : Any = vocab_size
# Backward compatibility with n_embed kwarg
a__ : Dict = kwargs.pop("n_embed" , lowerCamelCase__ )
a__ : str = hidden_size if n_embed is None else n_embed
a__ : Tuple = num_hidden_layers
a__ : Dict = num_attention_heads
a__ : Tuple = layer_norm_epsilon
a__ : List[str] = initializer_range
a__ : Dict = use_cache
a__ : Tuple = hidden_dropout
a__ : str = attention_dropout
a__ : str = bos_token_id
a__ : Any = eos_token_id
a__ : Tuple = num_attention_heads if num_kv_heads is None else num_kv_heads
a__ : Union[str, Any] = alibi
a__ : List[str] = new_decoder_architecture
a__ : Optional[Any] = multi_query # Ignored when new_decoder_architecture is True
a__ : int = parallel_attn
a__ : Any = bias
super().__init__(bos_token_id=lowerCamelCase__ , eos_token_id=lowerCamelCase__ , **lowerCamelCase__ )
@property
def _UpperCamelCase( self : Optional[Any] ):
return self.hidden_size // self.num_attention_heads
@property
def _UpperCamelCase( self : Dict ):
return not self.alibi
| 37 |
import argparse
import os
import numpy as np
import tensorflow as tf
import torch
from transformers import BertModel
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Union[str, Any] = ("dense.weight", "attention.self.query", "attention.self.key", "attention.self.value")
a__ : Union[str, Any] = (
("layer.", "layer_"),
("word_embeddings.weight", "word_embeddings"),
("position_embeddings.weight", "position_embeddings"),
("token_type_embeddings.weight", "token_type_embeddings"),
(".", "/"),
("LayerNorm/weight", "LayerNorm/gamma"),
("LayerNorm/bias", "LayerNorm/beta"),
("weight", "kernel"),
)
if not os.path.isdir(__a ):
os.makedirs(__a )
a__ : Any = model.state_dict()
def to_tf_var_name(__a ):
for patt, repl in iter(__a ):
a__ : Tuple = name.replace(__a , __a )
return f'''bert/{name}'''
def create_tf_var(__a , __a , __a ):
a__ : Tuple = tf.dtypes.as_dtype(tensor.dtype )
a__ : Dict = tf.get_variable(dtype=__a , shape=tensor.shape , name=__a , initializer=tf.zeros_initializer() )
session.run(tf.variables_initializer([tf_var] ) )
session.run(__a )
return tf_var
tf.reset_default_graph()
with tf.Session() as session:
for var_name in state_dict:
a__ : int = to_tf_var_name(__a )
a__ : Union[str, Any] = state_dict[var_name].numpy()
if any(x in var_name for x in tensors_to_transpose ):
a__ : int = torch_tensor.T
a__ : Optional[Any] = create_tf_var(tensor=__a , name=__a , session=__a )
tf.keras.backend.set_value(__a , __a )
a__ : int = session.run(__a )
print(f'''Successfully created {tf_name}: {np.allclose(__a , __a )}''' )
a__ : Any = tf.train.Saver(tf.trainable_variables() )
saver.save(__a , os.path.join(__a , model_name.replace("-" , "_" ) + ".ckpt" ) )
def UpperCamelCase_ ( __a=None ) -> int:
a__ : Dict = argparse.ArgumentParser()
parser.add_argument("--model_name" , type=__a , required=__a , help="model name e.g. bert-base-uncased" )
parser.add_argument(
"--cache_dir" , type=__a , default=__a , required=__a , help="Directory containing pytorch model" )
parser.add_argument("--pytorch_model_path" , type=__a , required=__a , help="/path/to/<pytorch-model-name>.bin" )
parser.add_argument("--tf_cache_dir" , type=__a , required=__a , help="Directory in which to save tensorflow model" )
a__ : Optional[Any] = parser.parse_args(__a )
a__ : Tuple = BertModel.from_pretrained(
pretrained_model_name_or_path=args.model_name , state_dict=torch.load(args.pytorch_model_path ) , cache_dir=args.cache_dir , )
convert_pytorch_checkpoint_to_tf(model=__a , ckpt_dir=args.tf_cache_dir , model_name=args.model_name )
if __name__ == "__main__":
main()
| 37 | 1 |
# Usage:
# ./gen-card-facebook-wmt19.py
import os
from pathlib import Path
def UpperCamelCase_ ( __a , __a , __a ) -> Any:
a__ : Optional[Any] = {
"en": "Machine learning is great, isn't it?",
"ru": "Машинное обучение - это здорово, не так ли?",
"de": "Maschinelles Lernen ist großartig, oder?",
}
# BLUE scores as follows:
# "pair": [fairseq, transformers]
a__ : Dict = {
"ru-en": ["[41.3](http://matrix.statmt.org/matrix/output/1907?run_id=6937)", "39.20"],
"en-ru": ["[36.4](http://matrix.statmt.org/matrix/output/1914?run_id=6724)", "33.47"],
"en-de": ["[43.1](http://matrix.statmt.org/matrix/output/1909?run_id=6862)", "42.83"],
"de-en": ["[42.3](http://matrix.statmt.org/matrix/output/1902?run_id=6750)", "41.35"],
}
a__ : Tuple = f'''{src_lang}-{tgt_lang}'''
a__ : int = f'''
---
language:
- {src_lang}
- {tgt_lang}
thumbnail:
tags:
- translation
- wmt19
- facebook
license: apache-2.0
datasets:
- wmt19
metrics:
- bleu
---
# FSMT
## Model description
This is a ported version of [fairseq wmt19 transformer](https://github.com/pytorch/fairseq/blob/master/examples/wmt19/README.md) for {src_lang}-{tgt_lang}.
For more details, please see, [Facebook FAIR\'s WMT19 News Translation Task Submission](https://arxiv.org/abs/1907.06616).
The abbreviation FSMT stands for FairSeqMachineTranslation
All four models are available:
* [wmt19-en-ru](https://huggingface.co/facebook/wmt19-en-ru)
* [wmt19-ru-en](https://huggingface.co/facebook/wmt19-ru-en)
* [wmt19-en-de](https://huggingface.co/facebook/wmt19-en-de)
* [wmt19-de-en](https://huggingface.co/facebook/wmt19-de-en)
## Intended uses & limitations
#### How to use
```python
from transformers import FSMTForConditionalGeneration, FSMTTokenizer
mname = "facebook/wmt19-{src_lang}-{tgt_lang}"
tokenizer = FSMTTokenizer.from_pretrained(mname)
model = FSMTForConditionalGeneration.from_pretrained(mname)
input = "{texts[src_lang]}"
input_ids = tokenizer.encode(input, return_tensors="pt")
outputs = model.generate(input_ids)
decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(decoded) # {texts[tgt_lang]}
```
#### Limitations and bias
- The original (and this ported model) doesn\'t seem to handle well inputs with repeated sub-phrases, [content gets truncated](https://discuss.huggingface.co/t/issues-with-translating-inputs-containing-repeated-phrases/981)
## Training data
Pretrained weights were left identical to the original model released by fairseq. For more details, please, see the [paper](https://arxiv.org/abs/1907.06616).
## Eval results
pair | fairseq | transformers
-------|---------|----------
{pair} | {scores[pair][0]} | {scores[pair][1]}
The score is slightly below the score reported by `fairseq`, since `transformers`` currently doesn\'t support:
- model ensemble, therefore the best performing checkpoint was ported (``model4.pt``).
- re-ranking
The score was calculated using this code:
```bash
git clone https://github.com/huggingface/transformers
cd transformers
export PAIR={pair}
export DATA_DIR=data/$PAIR
export SAVE_DIR=data/$PAIR
export BS=8
export NUM_BEAMS=15
mkdir -p $DATA_DIR
sacrebleu -t wmt19 -l $PAIR --echo src > $DATA_DIR/val.source
sacrebleu -t wmt19 -l $PAIR --echo ref > $DATA_DIR/val.target
echo $PAIR
PYTHONPATH="src:examples/seq2seq" python examples/seq2seq/run_eval.py facebook/wmt19-$PAIR $DATA_DIR/val.source $SAVE_DIR/test_translations.txt --reference_path $DATA_DIR/val.target --score_path $SAVE_DIR/test_bleu.json --bs $BS --task translation --num_beams $NUM_BEAMS
```
note: fairseq reports using a beam of 50, so you should get a slightly higher score if re-run with `--num_beams 50`.
## Data Sources
- [training, etc.](http://www.statmt.org/wmt19/)
- [test set](http://matrix.statmt.org/test_sets/newstest2019.tgz?1556572561)
### BibTeX entry and citation info
```bibtex
@inproceedings{{...,
year={{2020}},
title={{Facebook FAIR\'s WMT19 News Translation Task Submission}},
author={{Ng, Nathan and Yee, Kyra and Baevski, Alexei and Ott, Myle and Auli, Michael and Edunov, Sergey}},
booktitle={{Proc. of WMT}},
}}
```
## TODO
- port model ensemble (fairseq uses 4 model checkpoints)
'''
os.makedirs(__a , exist_ok=__a )
a__ : Union[str, Any] = os.path.join(__a , "README.md" )
print(f'''Generating {path}''' )
with open(__a , "w" , encoding="utf-8" ) as f:
f.write(__a )
# make sure we are under the root of the project
UpperCamelCase : Tuple = Path(__file__).resolve().parent.parent.parent
UpperCamelCase : Optional[int] = repo_dir / """model_cards"""
for model_name in ["wmt19-ru-en", "wmt19-en-ru", "wmt19-en-de", "wmt19-de-en"]:
UpperCamelCase , UpperCamelCase , UpperCamelCase : Optional[Any] = model_name.split("""-""")
UpperCamelCase : Dict = model_cards_dir / """facebook""" / model_name
write_model_card(model_card_dir, src_lang=src_lang, tgt_lang=tgt_lang)
| 37 |
import inspect
import unittest
from huggingface_hub import hf_hub_download
from transformers import ASTConfig
from transformers.testing_utils import require_torch, require_torchaudio, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_torchaudio_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import ASTForAudioClassification, ASTModel
from transformers.models.audio_spectrogram_transformer.modeling_audio_spectrogram_transformer import (
AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
)
if is_torchaudio_available():
import torchaudio
from transformers import ASTFeatureExtractor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[str] , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[Any]=2 , lowerCamelCase__ : Any=24 , lowerCamelCase__ : Optional[Any]=16 , lowerCamelCase__ : int=True , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[Any]=32 , lowerCamelCase__ : List[str]=5 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Optional[Any]=37 , lowerCamelCase__ : Any="gelu" , lowerCamelCase__ : Union[str, Any]=0.1 , lowerCamelCase__ : Optional[int]=0.1 , lowerCamelCase__ : str=10 , lowerCamelCase__ : Optional[Any]=0.02 , lowerCamelCase__ : str=None , lowerCamelCase__ : List[str]=2 , lowerCamelCase__ : Optional[Any]=2 , ):
a__ : str = parent
a__ : Any = batch_size
a__ : Dict = patch_size
a__ : List[Any] = max_length
a__ : str = num_mel_bins
a__ : Optional[Any] = is_training
a__ : Optional[int] = use_labels
a__ : List[Any] = hidden_size
a__ : str = num_hidden_layers
a__ : Any = num_attention_heads
a__ : Union[str, Any] = intermediate_size
a__ : List[str] = hidden_act
a__ : str = hidden_dropout_prob
a__ : Tuple = attention_probs_dropout_prob
a__ : List[Any] = type_sequence_label_size
a__ : Any = initializer_range
a__ : str = scope
a__ : List[str] = frequency_stride
a__ : Union[str, Any] = time_stride
# in AST, the seq length equals the number of patches + 2 (we add 2 for the [CLS] and distillation tokens)
a__ : List[Any] = (self.num_mel_bins - self.patch_size) // self.frequency_stride + 1
a__ : List[str] = (self.max_length - self.patch_size) // self.time_stride + 1
a__ : Tuple = frequency_out_dimension * time_out_dimension
a__ : List[str] = num_patches + 2
def _UpperCamelCase( self : List[str] ):
a__ : Any = floats_tensor([self.batch_size, self.max_length, self.num_mel_bins] )
a__ : List[Any] = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : List[str] = self.get_config()
return config, input_values, labels
def _UpperCamelCase( self : Optional[int] ):
return ASTConfig(
patch_size=self.patch_size , max_length=self.max_length , num_mel_bins=self.num_mel_bins , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , frequency_stride=self.frequency_stride , time_stride=self.time_stride , )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Tuple , lowerCamelCase__ : int , lowerCamelCase__ : Optional[int] ):
a__ : List[Any] = ASTModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Dict = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : str ):
a__ : Dict = self.prepare_config_and_inputs()
(
(
a__
), (
a__
), (
a__
),
) : Optional[int] = config_and_inputs
a__ : List[Any] = {"input_values": input_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(
ASTModel,
ASTForAudioClassification,
)
if is_torch_available()
else ()
)
_lowercase = (
{'audio-classification': ASTForAudioClassification, 'feature-extraction': ASTModel}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
if pipeline_test_casse_name == "AudioClassificationPipelineTests":
return True
return False
def _UpperCamelCase( self : str ):
a__ : str = ASTModelTester(self )
a__ : Any = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[str] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="AST does not use inputs_embeds" )
def _UpperCamelCase( self : List[str] ):
pass
def _UpperCamelCase( self : Optional[int] ):
a__, a__ : str = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Any = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Union[str, Any] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : Tuple ):
a__, a__ : List[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : Dict = model_class(lowerCamelCase__ )
a__ : Optional[int] = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Optional[Any] = ["input_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
for model_name in AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Union[str, Any] = ASTModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : Optional[int] = hf_hub_download(
repo_id="nielsr/audio-spectogram-transformer-checkpoint" , filename="sample_audio.flac" , repo_type="dataset" )
a__, a__ : List[str] = torchaudio.load(__a )
return audio, sampling_rate
@require_torch
@require_torchaudio
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : List[str] ):
return (
ASTFeatureExtractor.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" )
if is_torchaudio_available()
else None
)
@slow
def _UpperCamelCase( self : Optional[int] ):
a__ : int = self.default_feature_extractor
a__ : Optional[Any] = ASTForAudioClassification.from_pretrained("MIT/ast-finetuned-audioset-10-10-0.4593" ).to(lowerCamelCase__ )
a__ : Any = self.default_feature_extractor
a__, a__ : Dict = prepare_audio()
a__ : str = audio.squeeze().numpy()
a__ : Any = feature_extractor(lowerCamelCase__ , sampling_rate=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(**lowerCamelCase__ )
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 527) )
self.assertEqual(outputs.logits.shape , lowerCamelCase__ )
a__ : List[str] = torch.tensor([-0.8760, -7.0042, -8.6602] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(outputs.logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
| 37 | 1 |
def UpperCamelCase_ ( __a ) -> "list[int]":
if upper_limit < 0:
raise ValueError("Limit for the Catalan sequence must be ≥ 0" )
a__ : List[str] = [0] * (upper_limit + 1)
# Base case: C(0) = C(1) = 1
a__ : Optional[int] = 1
if upper_limit > 0:
a__ : Optional[Any] = 1
# Recurrence relation: C(i) = sum(C(j).C(i-j-1)), from j = 0 to i
for i in range(2 , upper_limit + 1 ):
for j in range(__a ):
catalan_list[i] += catalan_list[j] * catalan_list[i - j - 1]
return catalan_list
if __name__ == "__main__":
print("""\n********* Catalan Numbers Using Dynamic Programming ************\n""")
print("""\n*** Enter -1 at any time to quit ***""")
print("""\nEnter the upper limit (≥ 0) for the Catalan number sequence: """, end="""""")
try:
while True:
UpperCamelCase : Optional[Any] = int(input().strip())
if N < 0:
print("""\n********* Goodbye!! ************""")
break
else:
print(f"""The Catalan numbers from 0 through {N} are:""")
print(catalan_numbers(N))
print("""Try another upper limit for the sequence: """, end="""""")
except (NameError, ValueError):
print("""\n********* Invalid input, goodbye! ************\n""")
import doctest
doctest.testmod()
| 37 |
import pickle
import shutil
import tempfile
import unittest
from transformers import SPIECE_UNDERLINE, XGLMTokenizer, XGLMTokenizerFast
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers, slow
from transformers.utils import cached_property
from ...test_tokenization_common import TokenizerTesterMixin
UpperCamelCase : Optional[Any] = get_tests_dir("""fixtures/test_sentencepiece.model""")
@require_sentencepiece
@require_tokenizers
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = XGLMTokenizer
_lowercase = XGLMTokenizerFast
_lowercase = True
_lowercase = True
def _UpperCamelCase( self : List[Any] ):
super().setUp()
# We have a SentencePiece fixture for testing
a__ : str = XGLMTokenizer(lowerCamelCase__ , keep_accents=lowerCamelCase__ )
tokenizer.save_pretrained(self.tmpdirname )
def _UpperCamelCase( self : List[Any] ):
a__ : int = "<pad>"
a__ : Union[str, Any] = 1
self.assertEqual(self.get_tokenizer()._convert_token_to_id(lowerCamelCase__ ) , lowerCamelCase__ )
self.assertEqual(self.get_tokenizer()._convert_id_to_token(lowerCamelCase__ ) , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
a__ : List[str] = list(self.get_tokenizer().get_vocab().keys() )
self.assertEqual(vocab_keys[0] , "<s>" )
self.assertEqual(vocab_keys[1] , "<pad>" )
self.assertEqual(len(lowerCamelCase__ ) , 1_008 )
def _UpperCamelCase( self : Dict ):
self.assertEqual(self.get_tokenizer().vocab_size , 1_008 )
def _UpperCamelCase( self : Optional[int] ):
a__ : str = XGLMTokenizer(lowerCamelCase__ , keep_accents=lowerCamelCase__ )
a__ : List[str] = tokenizer.tokenize("This is a test" )
self.assertListEqual(lowerCamelCase__ , ["▁This", "▁is", "▁a", "▁t", "est"] )
self.assertListEqual(
tokenizer.convert_tokens_to_ids(lowerCamelCase__ ) , [value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]] , )
a__ : Any = tokenizer.tokenize("I was born in 92000, and this is falsé." )
self.assertListEqual(
lowerCamelCase__ , [
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"9",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"é",
".",
] , )
a__ : List[str] = tokenizer.convert_tokens_to_ids(lowerCamelCase__ )
self.assertListEqual(
lowerCamelCase__ , [
value + tokenizer.fairseq_offset
for value in [8, 21, 84, 55, 24, 19, 7, 2, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, 2, 4]
] , )
a__ : Dict = tokenizer.convert_ids_to_tokens(lowerCamelCase__ )
self.assertListEqual(
lowerCamelCase__ , [
SPIECE_UNDERLINE + "I",
SPIECE_UNDERLINE + "was",
SPIECE_UNDERLINE + "b",
"or",
"n",
SPIECE_UNDERLINE + "in",
SPIECE_UNDERLINE + "",
"<unk>",
"2",
"0",
"0",
"0",
",",
SPIECE_UNDERLINE + "and",
SPIECE_UNDERLINE + "this",
SPIECE_UNDERLINE + "is",
SPIECE_UNDERLINE + "f",
"al",
"s",
"<unk>",
".",
] , )
@cached_property
def _UpperCamelCase( self : Dict ):
return XGLMTokenizer.from_pretrained("facebook/xglm-564M" )
def _UpperCamelCase( self : Union[str, Any] ):
with tempfile.NamedTemporaryFile() as f:
shutil.copyfile(lowerCamelCase__ , f.name )
a__ : Any = XGLMTokenizer(f.name , keep_accents=lowerCamelCase__ )
a__ : List[str] = pickle.dumps(lowerCamelCase__ )
pickle.loads(lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
if not self.test_rust_tokenizer:
return
a__ : Any = self.get_tokenizer()
a__ : Optional[Any] = self.get_rust_tokenizer()
a__ : Tuple = "I was born in 92000, and this is falsé."
a__ : List[str] = tokenizer.tokenize(lowerCamelCase__ )
a__ : Union[str, Any] = rust_tokenizer.tokenize(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : Optional[int] = tokenizer.encode(lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
a__ : Union[str, Any] = rust_tokenizer.encode(lowerCamelCase__ , add_special_tokens=lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
a__ : List[str] = self.get_rust_tokenizer()
a__ : Tuple = tokenizer.encode(lowerCamelCase__ )
a__ : Optional[Any] = rust_tokenizer.encode(lowerCamelCase__ )
self.assertListEqual(lowerCamelCase__ , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : List[str] ):
a__ : Union[str, Any] = "Hello World!"
a__ : List[str] = [2, 31_227, 4_447, 35]
self.assertListEqual(lowerCamelCase__ , self.big_tokenizer.encode(lowerCamelCase__ ) )
@slow
def _UpperCamelCase( self : Union[str, Any] ):
a__ : Optional[int] = (
"This is a very long text with a lot of weird characters, such as: . , ~ ? ( ) \" [ ] ! : - . Also we will"
" add words that should not exsist and be tokenized to unk, such as saoneuhaoesuth"
)
# fmt: off
a__ : Union[str, Any] = [2, 1_018, 67, 11, 1_988, 2_617, 5_631, 278, 11, 3_407, 48, 71_630, 28_085, 4, 3_234, 157, 13, 6, 5, 6, 4, 3_526, 768, 15, 659, 57, 298, 3_983, 864, 129, 21, 6, 5, 13_675, 377, 652, 7_580, 10_341, 155, 2_817, 422, 1_666, 7, 1_674, 53, 113, 202_277, 17_892, 33, 60, 87, 4, 3_234, 157, 61, 2_667, 52_376, 19, 88, 23, 735]
# fmt: on
self.assertListEqual(lowerCamelCase__ , self.big_tokenizer.encode(lowerCamelCase__ ) )
@slow
def _UpperCamelCase( self : List[Any] ):
# fmt: off
a__ : Optional[int] = {
"input_ids": [[2, 108_825, 1_163, 15, 88_010, 473, 15_898, 157, 13_672, 1_857, 312, 8, 238_021, 1_163, 53, 13_672, 1_857, 312, 8, 53_283, 182_396, 8, 18_566, 16, 36_733, 4_101, 8, 230, 244_017, 122_553, 7, 15, 132_597, 4, 293, 12_511, 7_610, 4, 3_414, 132_597, 9, 4, 32_361, 362, 4, 734, 28_512, 32_569, 18, 4, 32_361, 26_096, 14_982, 73, 18_715, 21_433, 235_261, 15, 492, 12_427, 16, 53, 18_715, 21_433, 65_454, 15, 23_659, 563, 16, 278, 597, 2_843, 595, 7_931, 182_396, 64_186, 22, 886, 595, 132_981, 53, 25_540, 3_449, 43_982, 39_901, 5_951, 878, 330, 4, 27_694, 80_269, 312, 53, 6_517, 11_780, 611, 20_408, 5], [2, 6, 132_597, 67, 42_897, 33, 592, 8, 163_729, 25_540, 361, 136_997, 109_514, 173_230, 7, 501, 60, 102_913, 196, 5_631, 235, 63_243, 473, 6, 231_757, 74, 5_277, 7_905, 53, 3_095, 37_317, 22, 454, 183_874, 5], [2, 268, 31_298, 46_530, 6, 132_935, 43_831, 7, 597, 32, 24, 3_688, 9_865, 5]],
"attention_mask": [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]
} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=lowerCamelCase__ , model_name="facebook/xglm-564M" , padding=lowerCamelCase__ , )
| 37 | 1 |
import shutil
import tempfile
import unittest
from unittest.mock import patch
from transformers import (
DefaultFlowCallback,
IntervalStrategy,
PrinterCallback,
ProgressCallback,
Trainer,
TrainerCallback,
TrainingArguments,
is_torch_available,
)
from transformers.testing_utils import require_torch
if is_torch_available():
from transformers.trainer import DEFAULT_CALLBACKS
from .test_trainer import RegressionDataset, RegressionModelConfig, RegressionPreTrainedModel
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Any ):
a__ : List[Any] = []
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Dict , **lowerCamelCase__ : Optional[Any] ):
self.events.append("on_init_end" )
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Dict , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Any , **lowerCamelCase__ : Optional[int] ):
self.events.append("on_train_begin" )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : str , **lowerCamelCase__ : Tuple ):
self.events.append("on_train_end" )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : int , lowerCamelCase__ : str , lowerCamelCase__ : Any , **lowerCamelCase__ : Optional[int] ):
self.events.append("on_epoch_begin" )
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : List[Any] , **lowerCamelCase__ : Union[str, Any] ):
self.events.append("on_epoch_end" )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : int ):
self.events.append("on_step_begin" )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[Any] , **lowerCamelCase__ : Optional[Any] ):
self.events.append("on_step_end" )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Any , **lowerCamelCase__ : int ):
self.events.append("on_evaluate" )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str , lowerCamelCase__ : List[str] , lowerCamelCase__ : List[str] , **lowerCamelCase__ : Dict ):
self.events.append("on_predict" )
def _UpperCamelCase( self : int , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : str , **lowerCamelCase__ : int ):
self.events.append("on_save" )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : List[Any] , lowerCamelCase__ : List[Any] , **lowerCamelCase__ : Optional[int] ):
self.events.append("on_log" )
def _UpperCamelCase( self : Any , lowerCamelCase__ : int , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Tuple , **lowerCamelCase__ : List[str] ):
self.events.append("on_prediction_step" )
@require_torch
class A__ ( unittest.TestCase ):
"""simple docstring"""
def _UpperCamelCase( self : Optional[Any] ):
a__ : str = tempfile.mkdtemp()
def _UpperCamelCase( self : str ):
shutil.rmtree(self.output_dir )
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : Optional[int]=0 , lowerCamelCase__ : Optional[Any]=0 , lowerCamelCase__ : Optional[Any]=64 , lowerCamelCase__ : Union[str, Any]=64 , lowerCamelCase__ : Optional[int]=None , lowerCamelCase__ : Dict=False , **lowerCamelCase__ : Optional[Any] ):
# disable_tqdm in TrainingArguments has a flaky default since it depends on the level of logging. We make sure
# its set to False since the tests later on depend on its value.
a__ : Optional[Any] = RegressionDataset(length=lowerCamelCase__ )
a__ : Union[str, Any] = RegressionDataset(length=lowerCamelCase__ )
a__ : Dict = RegressionModelConfig(a=lowerCamelCase__ , b=lowerCamelCase__ )
a__ : List[str] = RegressionPreTrainedModel(lowerCamelCase__ )
a__ : int = TrainingArguments(self.output_dir , disable_tqdm=lowerCamelCase__ , report_to=[] , **lowerCamelCase__ )
return Trainer(
lowerCamelCase__ , lowerCamelCase__ , train_dataset=lowerCamelCase__ , eval_dataset=lowerCamelCase__ , callbacks=lowerCamelCase__ , )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : List[Any] ):
self.assertEqual(len(lowerCamelCase__ ) , len(lowerCamelCase__ ) )
# Order doesn't matter
a__ : Union[str, Any] = sorted(lowerCamelCase__ , key=lambda lowerCamelCase__ : cb.__name__ if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else cb.__class__.__name__ )
a__ : Optional[Any] = sorted(lowerCamelCase__ , key=lambda lowerCamelCase__ : cb.__name__ if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else cb.__class__.__name__ )
for cba, cba in zip(lowerCamelCase__ , lowerCamelCase__ ):
if isinstance(lowerCamelCase__ , lowerCamelCase__ ) and isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertEqual(lowerCamelCase__ , lowerCamelCase__ )
elif isinstance(lowerCamelCase__ , lowerCamelCase__ ) and not isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertEqual(lowerCamelCase__ , cba.__class__ )
elif not isinstance(lowerCamelCase__ , lowerCamelCase__ ) and isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertEqual(cba.__class__ , lowerCamelCase__ )
else:
self.assertEqual(lowerCamelCase__ , lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : str ):
a__ : Optional[int] = ["on_init_end", "on_train_begin"]
a__ : Dict = 0
a__ : List[Any] = len(trainer.get_eval_dataloader() )
a__ : Dict = ["on_prediction_step"] * len(trainer.get_eval_dataloader() ) + ["on_log", "on_evaluate"]
for _ in range(trainer.state.num_train_epochs ):
expected_events.append("on_epoch_begin" )
for _ in range(lowerCamelCase__ ):
step += 1
expected_events += ["on_step_begin", "on_step_end"]
if step % trainer.args.logging_steps == 0:
expected_events.append("on_log" )
if trainer.args.evaluation_strategy == IntervalStrategy.STEPS and step % trainer.args.eval_steps == 0:
expected_events += evaluation_events.copy()
if step % trainer.args.save_steps == 0:
expected_events.append("on_save" )
expected_events.append("on_epoch_end" )
if trainer.args.evaluation_strategy == IntervalStrategy.EPOCH:
expected_events += evaluation_events.copy()
expected_events += ["on_log", "on_train_end"]
return expected_events
def _UpperCamelCase( self : str ):
a__ : str = self.get_trainer()
a__ : int = DEFAULT_CALLBACKS.copy() + [ProgressCallback]
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
# Callbacks passed at init are added to the default callbacks
a__ : Union[str, Any] = self.get_trainer(callbacks=[MyTestTrainerCallback] )
expected_callbacks.append(lowerCamelCase__ )
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
# TrainingArguments.disable_tqdm controls if use ProgressCallback or PrinterCallback
a__ : Any = self.get_trainer(disable_tqdm=lowerCamelCase__ )
a__ : Dict = DEFAULT_CALLBACKS.copy() + [PrinterCallback]
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
def _UpperCamelCase( self : Tuple ):
a__ : List[str] = DEFAULT_CALLBACKS.copy() + [ProgressCallback]
a__ : str = self.get_trainer()
# We can add, pop, or remove by class name
trainer.remove_callback(lowerCamelCase__ )
expected_callbacks.remove(lowerCamelCase__ )
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
a__ : List[Any] = self.get_trainer()
a__ : Dict = trainer.pop_callback(lowerCamelCase__ )
self.assertEqual(cb.__class__ , lowerCamelCase__ )
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
trainer.add_callback(lowerCamelCase__ )
expected_callbacks.insert(0 , lowerCamelCase__ )
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
# We can also add, pop, or remove by instance
a__ : Union[str, Any] = self.get_trainer()
a__ : Optional[int] = trainer.callback_handler.callbacks[0]
trainer.remove_callback(lowerCamelCase__ )
expected_callbacks.remove(lowerCamelCase__ )
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
a__ : Dict = self.get_trainer()
a__ : List[Any] = trainer.callback_handler.callbacks[0]
a__ : Optional[int] = trainer.pop_callback(lowerCamelCase__ )
self.assertEqual(lowerCamelCase__ , lowerCamelCase__ )
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
trainer.add_callback(lowerCamelCase__ )
expected_callbacks.insert(0 , lowerCamelCase__ )
self.check_callbacks_equality(trainer.callback_handler.callbacks , lowerCamelCase__ )
def _UpperCamelCase( self : str ):
import warnings
# XXX: for now ignore scatter_gather warnings in this test since it's not relevant to what's being tested
warnings.simplefilter(action="ignore" , category=lowerCamelCase__ )
a__ : List[str] = self.get_trainer(callbacks=[MyTestTrainerCallback] )
trainer.train()
a__ : List[str] = trainer.callback_handler.callbacks[-2].events
self.assertEqual(lowerCamelCase__ , self.get_expected_events(lowerCamelCase__ ) )
# Independent log/save/eval
a__ : int = self.get_trainer(callbacks=[MyTestTrainerCallback] , logging_steps=5 )
trainer.train()
a__ : List[str] = trainer.callback_handler.callbacks[-2].events
self.assertEqual(lowerCamelCase__ , self.get_expected_events(lowerCamelCase__ ) )
a__ : str = self.get_trainer(callbacks=[MyTestTrainerCallback] , save_steps=5 )
trainer.train()
a__ : str = trainer.callback_handler.callbacks[-2].events
self.assertEqual(lowerCamelCase__ , self.get_expected_events(lowerCamelCase__ ) )
a__ : str = self.get_trainer(callbacks=[MyTestTrainerCallback] , eval_steps=5 , evaluation_strategy="steps" )
trainer.train()
a__ : List[str] = trainer.callback_handler.callbacks[-2].events
self.assertEqual(lowerCamelCase__ , self.get_expected_events(lowerCamelCase__ ) )
a__ : Any = self.get_trainer(callbacks=[MyTestTrainerCallback] , evaluation_strategy="epoch" )
trainer.train()
a__ : int = trainer.callback_handler.callbacks[-2].events
self.assertEqual(lowerCamelCase__ , self.get_expected_events(lowerCamelCase__ ) )
# A bit of everything
a__ : List[Any] = self.get_trainer(
callbacks=[MyTestTrainerCallback] , logging_steps=3 , save_steps=10 , eval_steps=5 , evaluation_strategy="steps" , )
trainer.train()
a__ : str = trainer.callback_handler.callbacks[-2].events
self.assertEqual(lowerCamelCase__ , self.get_expected_events(lowerCamelCase__ ) )
# warning should be emitted for duplicated callbacks
with patch("transformers.trainer_callback.logger.warning" ) as warn_mock:
a__ : List[str] = self.get_trainer(
callbacks=[MyTestTrainerCallback, MyTestTrainerCallback] , )
assert str(lowerCamelCase__ ) in warn_mock.call_args[0][0]
| 37 |
import argparse
import requests
import torch
# pip3 install salesforce-lavis
# I'm actually installing a slightly modified version: pip3 install git+https://github.com/nielsrogge/LAVIS.git@fix_lavis_float32 (there's also the fix_lavis branch)
# also note: to convert Vicuna checkpoints, we had to include /home/niels/python_projects/checkpoints/FastChat/vicuna-7b in lavis/configs/models/blip2/blip2_instruct_vicuna7b.yaml
# same for Vicuna-13b
from lavis.models import load_model_and_preprocess
from PIL import Image
from transformers import (
AutoTokenizer,
BlipImageProcessor,
InstructBlipConfig,
InstructBlipForConditionalGeneration,
InstructBlipProcessor,
InstructBlipQFormerConfig,
InstructBlipVisionConfig,
LlamaConfig,
LlamaTokenizerFast,
TaConfig,
TaTokenizerFast,
)
from transformers.utils.constants import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD
def UpperCamelCase_ ( ) -> int:
a__ : int = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
a__ : Optional[Any] = Image.open(requests.get(__a , stream=__a ).raw ).convert("RGB" )
return image
def UpperCamelCase_ ( __a ) -> Optional[Any]:
a__ : Any = []
# fmt: off
# vision encoder
rename_keys.append(("visual_encoder.cls_token", "vision_model.embeddings.class_embedding") )
rename_keys.append(("visual_encoder.pos_embed", "vision_model.embeddings.position_embedding") )
rename_keys.append(("visual_encoder.patch_embed.proj.weight", "vision_model.embeddings.patch_embedding.weight") )
rename_keys.append(("visual_encoder.patch_embed.proj.bias", "vision_model.embeddings.patch_embedding.bias") )
rename_keys.append(("ln_vision.weight", "vision_model.post_layernorm.weight") )
rename_keys.append(("ln_vision.bias", "vision_model.post_layernorm.bias") )
for i in range(config.vision_config.num_hidden_layers ):
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.weight''', f'''vision_model.encoder.layers.{i}.layer_norm1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm1.bias''', f'''vision_model.encoder.layers.{i}.layer_norm1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.weight''', f'''vision_model.encoder.layers.{i}.layer_norm2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.norm2.bias''', f'''vision_model.encoder.layers.{i}.layer_norm2.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.qkv.weight''', f'''vision_model.encoder.layers.{i}.self_attn.qkv.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.weight''', f'''vision_model.encoder.layers.{i}.self_attn.projection.weight''',) )
rename_keys.append((f'''visual_encoder.blocks.{i}.attn.proj.bias''', f'''vision_model.encoder.layers.{i}.self_attn.projection.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc1.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc1.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc1.bias''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.weight''', f'''vision_model.encoder.layers.{i}.mlp.fc2.weight''') )
rename_keys.append((f'''visual_encoder.blocks.{i}.mlp.fc2.bias''', f'''vision_model.encoder.layers.{i}.mlp.fc2.bias''') )
# QFormer
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.weight", "qformer.embeddings.layernorm.weight") )
rename_keys.append(("Qformer.bert.embeddings.LayerNorm.bias", "qformer.embeddings.layernorm.bias") )
# fmt: on
return rename_keys
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Union[str, Any] = dct.pop(__a )
a__ : List[str] = val
def UpperCamelCase_ ( __a , __a ) -> Optional[Any]:
for i in range(config.vision_config.num_hidden_layers ):
# read in original q and v biases
a__ : Any = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.q_bias''' )
a__ : Tuple = state_dict.pop(f'''visual_encoder.blocks.{i}.attn.v_bias''' )
# next, set bias in the state dict
a__ : str = torch.cat((q_bias, torch.zeros_like(__a , requires_grad=__a ), v_bias) )
a__ : int = qkv_bias
def UpperCamelCase_ ( __a ) -> Dict:
a__ : Tuple = 364 if "coco" in model_name else 224
a__ : int = InstructBlipVisionConfig(image_size=__a ).to_dict()
# make sure the models have proper bos_token_id and eos_token_id set (important for generation)
# seems like flan-T5 models don't have bos_token_id properly set?
if "t5-xl" in model_name:
a__ : Tuple = TaConfig.from_pretrained("google/flan-t5-xl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "t5-xxl" in model_name:
a__ : Dict = TaConfig.from_pretrained("google/flan-t5-xxl" , dense_act_fn="gelu" , bos_token_id=1 ).to_dict()
elif "vicuna-7b" in model_name:
a__ : List[Any] = LlamaConfig.from_pretrained("decapoda-research/llama-7b-hf" , vocab_size=32_001 ).to_dict()
elif "vicuna-13b" in model_name:
a__ : Optional[int] = LlamaConfig.from_pretrained("decapoda-research/llama-13b-hf" , vocab_size=32_001 ).to_dict()
else:
raise ValueError("Model name not supported" )
# the authors add one special "[DEC]" token to the vocab of Q-Former, hence vocab size = 30522 + 1
a__ : Optional[Any] = InstructBlipQFormerConfig(vocab_size=30_523 ).to_dict()
a__ : Any = InstructBlipConfig(vision_config=__a , text_config=__a , qformer_config=__a )
return config, image_size
@torch.no_grad()
def UpperCamelCase_ ( __a , __a=None , __a=False ) -> int:
a__ : Tuple = AutoTokenizer.from_pretrained("bert-base-uncased" , truncation_side="left" )
qformer_tokenizer.add_special_tokens({"bos_token": "[DEC]"} )
if "t5" in model_name:
a__ : List[Any] = TaTokenizerFast.from_pretrained("google/flan-t5-xl" , truncation_side="left" )
elif "vicuna" in model_name:
# the following was used in the original implementation:
# tokenizer = LlamaTokenizer.from_pretrained("huggyllama/llama-7b", use_fast=False, truncation_side="left")
# tokenizer.add_special_tokens({"pad_token": "[PAD]"})
# tokenizer.add_special_tokens({"bos_token": "</s>"})
# tokenizer.add_special_tokens({"eos_token": "</s>"})
# tokenizer.add_special_tokens({"unk_token": "</s>"})
a__ : Union[str, Any] = LlamaTokenizerFast.from_pretrained(
"huggyllama/llama-7b" , truncation_side="left" , bos_token="</s>" , unk_token="</s>" )
tokenizer.add_special_tokens({"pad_token": "[PAD]"} )
a__, a__ : List[str] = get_blipa_config(__a )
a__ : Any = InstructBlipForConditionalGeneration(__a ).eval()
a__ : Dict = {
"instructblip-vicuna-7b": ("blip2_vicuna_instruct", "vicuna7b"),
"instructblip-vicuna-13b": ("blip2_vicuna_instruct", "vicuna13b"),
"instructblip-flan-t5-xl": ("blip2_t5_instruct", "flant5xl"),
"instructblip-flan-t5-xxl": ("blip2_t5_instruct", "flant5xxl"),
}
a__, a__ : Dict = model_name_to_original[model_name]
# load original model
print("Loading original model..." )
a__ : Optional[Any] = "cuda:1" if torch.cuda.is_available() else "cpu"
a__ : List[Any] = "cuda:2" if torch.cuda.is_available() else "cpu"
a__, a__, a__ : Tuple = load_model_and_preprocess(
name=__a , model_type=__a , is_eval=__a , device=__a )
original_model.eval()
print("Done!" )
# update state dict keys
a__ : Dict = original_model.state_dict()
a__ : Optional[int] = create_rename_keys(__a )
for src, dest in rename_keys:
rename_key(__a , __a , __a )
# some keys can be renamed efficiently
for key, val in state_dict.copy().items():
a__ : Optional[int] = state_dict.pop(__a )
if key.startswith("Qformer.bert" ):
a__ : List[Any] = key.replace("Qformer.bert" , "qformer" )
if "attention.self" in key:
a__ : Any = key.replace("self" , "attention" )
if "llm_proj" in key:
a__ : Dict = key.replace("llm_proj" , "language_projection" )
if "t5_proj" in key:
a__ : int = key.replace("t5_proj" , "language_projection" )
if key.startswith("llm_model" ):
a__ : List[str] = key.replace("llm_model" , "language_model" )
if key.startswith("t5" ):
a__ : str = key.replace("t5" , "language" )
a__ : Dict = val
# read in qv biases
read_in_q_v_bias(__a , __a )
# note: weights get loaded in torch.float32 by default
hf_model.load_state_dict(__a , strict=__a )
a__ : Union[str, Any] = load_demo_image()
a__ : int = "What is unusual about this image?"
# create processor
a__ : Any = BlipImageProcessor(
size={"height": image_size, "width": image_size} , image_mean=__a , image_std=__a )
a__ : Tuple = InstructBlipProcessor(
image_processor=__a , tokenizer=__a , qformer_tokenizer=__a , )
a__ : Tuple = processor(images=__a , text=__a , return_tensors="pt" ).to(__a )
# make sure processor creates exact same pixel values
a__ : Optional[int] = vis_processors["eval"](__a ).unsqueeze(0 ).to(__a )
a__ : Optional[Any] = inputs.pixel_values
assert torch.allclose(original_pixel_values.to(pixel_values.device ) , __a )
original_model.to(__a )
hf_model.to(__a )
with torch.no_grad():
if "vicuna" in model_name:
a__ : str = original_model({"image": original_pixel_values, "text_input": [prompt]} ).logits
a__ : List[str] = hf_model(**__a ).logits
else:
a__ : List[Any] = original_model(
{"image": original_pixel_values, "text_input": [prompt], "text_output": ["\n"]} ).logits
a__ : str = tokenizer("\n" , return_tensors="pt" ).input_ids.to(__a )
a__ : Dict = label_input_ids.masked_fill(label_input_ids == tokenizer.pad_token_id , -100 )
a__ : Any = hf_model(**__a , labels=__a ).logits
print("First values of original logits:" , original_logits[0, :3, :3] )
print("First values of HF logits:" , logits[0, :3, :3] )
# assert values
assert original_logits.shape == logits.shape
a__ : Tuple = 1e-4 if "vicuna" in model_name else 1e-5
assert torch.allclose(original_logits.to(logits.device ) , __a , atol=__a )
print("Looks ok!" )
print("Generating with original model..." )
a__ : Tuple = original_model.generate({"image": original_pixel_values, "prompt": prompt} , num_beams=5 )
# important: we need to cast the weights of the HF model to the appropriate type
print("Generating with HF model..." )
a__ : int = hf_model.generate(
**__a , do_sample=__a , num_beams=5 , max_length=256 , min_length=1 , top_p=0.9 , repetition_penalty=1.5 , length_penalty=1.0 , temperature=1 , )
if "vicuna" in model_name:
# convert output id 0 to 2 (eos_token_id)
# TODO add this in the generate method?
a__ : int = 2
print("Original generation:" , __a )
a__ : str = processor.batch_decode(__a , skip_special_tokens=__a )
a__ : str = [text.strip() for text in output_text]
print("HF generation:" , __a )
if pytorch_dump_folder_path is not None:
processor.save_pretrained(__a )
hf_model.save_pretrained(__a )
if push_to_hub:
processor.push_to_hub(f'''Salesforce/{model_name}''' )
hf_model.push_to_hub(f'''Salesforce/{model_name}''' )
if __name__ == "__main__":
UpperCamelCase : Any = argparse.ArgumentParser()
UpperCamelCase : Optional[int] = [
"""instructblip-vicuna-7b""",
"""instructblip-vicuna-13b""",
"""instructblip-flan-t5-xl""",
"""instructblip-flan-t5-xxl""",
]
parser.add_argument(
"""--model_name""",
default="""instructblip-flan-t5-xl""",
choices=choices,
type=str,
help="""Path to hf config.json of model to convert""",
)
parser.add_argument("""--pytorch_dump_folder_path""", default=None, type=str, help="""Path to the output PyTorch model.""")
parser.add_argument(
"""--push_to_hub""",
action="""store_true""",
help="""Whether to push the model and processor to the hub after converting""",
)
UpperCamelCase : Dict = parser.parse_args()
convert_blipa_checkpoint(args.model_name, args.pytorch_dump_folder_path, args.push_to_hub)
| 37 | 1 |
import copy
import random
from transformers import CLIPTokenizer
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Dict , *lowerCamelCase__ : Tuple , **lowerCamelCase__ : Optional[Any] ):
super().__init__(*lowerCamelCase__ , **lowerCamelCase__ )
a__ : Dict = {}
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Dict , *lowerCamelCase__ : int , **lowerCamelCase__ : List[str] ):
a__ : Optional[Any] = super().add_tokens(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ )
if num_added_tokens == 0:
raise ValueError(
f'''The tokenizer already contains the token {placeholder_token}. Please pass a different'''
" `placeholder_token` that is not already in the tokenizer." )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Dict , *lowerCamelCase__ : Optional[Any] , lowerCamelCase__ : Any=1 , **lowerCamelCase__ : List[str] ):
a__ : int = []
if num_vec_per_token == 1:
self.try_adding_tokens(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ )
output.append(lowerCamelCase__ )
else:
a__ : str = []
for i in range(lowerCamelCase__ ):
a__ : List[str] = placeholder_token + f'''_{i}'''
self.try_adding_tokens(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ )
output.append(lowerCamelCase__ )
# handle cases where there is a new placeholder token that contains the current placeholder token but is larger
for token in self.token_map:
if token in placeholder_token:
raise ValueError(
f'''The tokenizer already has placeholder token {token} that can get confused with'''
f''' {placeholder_token}keep placeholder tokens independent''' )
a__ : List[Any] = output
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : List[str] , lowerCamelCase__ : Tuple=False , lowerCamelCase__ : int=1.0 ):
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
a__ : Any = []
for i in range(len(lowerCamelCase__ ) ):
output.append(self.replace_placeholder_tokens_in_text(text[i] , vector_shuffle=lowerCamelCase__ ) )
return output
for placeholder_token in self.token_map:
if placeholder_token in text:
a__ : Any = self.token_map[placeholder_token]
a__ : Optional[Any] = tokens[: 1 + int(len(lowerCamelCase__ ) * prop_tokens_to_load )]
if vector_shuffle:
a__ : str = copy.copy(lowerCamelCase__ )
random.shuffle(lowerCamelCase__ )
a__ : Union[str, Any] = text.replace(lowerCamelCase__ , " ".join(lowerCamelCase__ ) )
return text
def __call__( self : str , lowerCamelCase__ : Union[str, Any] , *lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : Optional[int]=False , lowerCamelCase__ : List[str]=1.0 , **lowerCamelCase__ : Tuple ):
return super().__call__(
self.replace_placeholder_tokens_in_text(
lowerCamelCase__ , vector_shuffle=lowerCamelCase__ , prop_tokens_to_load=lowerCamelCase__ ) , *lowerCamelCase__ , **lowerCamelCase__ , )
def _UpperCamelCase( self : Any , lowerCamelCase__ : Optional[int] , *lowerCamelCase__ : int , lowerCamelCase__ : Union[str, Any]=False , lowerCamelCase__ : Any=1.0 , **lowerCamelCase__ : List[str] ):
return super().encode(
self.replace_placeholder_tokens_in_text(
lowerCamelCase__ , vector_shuffle=lowerCamelCase__ , prop_tokens_to_load=lowerCamelCase__ ) , *lowerCamelCase__ , **lowerCamelCase__ , )
| 37 |
def UpperCamelCase_ ( __a , __a ) -> Tuple:
a__ : Optional[int] = [0 for i in range(r + 1 )]
# nc0 = 1
a__ : Union[str, Any] = 1
for i in range(1 , n + 1 ):
# to compute current row from previous row.
a__ : Any = min(__a , __a )
while j > 0:
c[j] += c[j - 1]
j -= 1
return c[r]
print(binomial_coefficient(n=10, r=5))
| 37 | 1 |
import os
from typing import Optional
import fsspec
from fsspec.archive import AbstractArchiveFileSystem
from fsspec.utils import DEFAULT_BLOCK_SIZE
class A__ ( A__ ):
"""simple docstring"""
_lowercase = ''
_lowercase = (
None # protocol passed in prefix to the url. ex: "gzip", for gzip://file.txt::http://foo.bar/file.txt.gz
)
_lowercase = None # compression type in fsspec. ex: "gzip"
_lowercase = None # extension of the filename to strip. ex: "".gz" to get file.txt from file.txt.gz
def __init__( self : List[str] , lowerCamelCase__ : str = "" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , **lowerCamelCase__ : List[str] ):
super().__init__(self , **lowerCamelCase__ )
# always open as "rb" since fsspec can then use the TextIOWrapper to make it work for "r" mode
a__ : str = fsspec.open(
lowerCamelCase__ , mode="rb" , protocol=lowerCamelCase__ , compression=self.compression , client_kwargs={
"requote_redirect_url": False, # see https://github.com/huggingface/datasets/pull/5459
"trust_env": True, # Enable reading proxy env variables.
**(target_options or {}).pop("client_kwargs" , {} ), # To avoid issues if it was already passed.
} , **(target_options or {}) , )
a__ : Optional[int] = os.path.basename(self.file.path.split("::" )[0] )
a__ : int = (
self.compressed_name[: self.compressed_name.rindex("." )]
if "." in self.compressed_name
else self.compressed_name
)
a__ : List[Any] = None
@classmethod
def _UpperCamelCase( cls : int , lowerCamelCase__ : int ):
# compressed file paths are always relative to the archive root
return super()._strip_protocol(lowerCamelCase__ ).lstrip("/" )
def _UpperCamelCase( self : Dict ):
if self.dir_cache is None:
a__ : Dict = {**self.file.fs.info(self.file.path ), "name": self.uncompressed_name}
a__ : int = {f["name"]: f}
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str ):
return self.file.open().read()
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : int=None , lowerCamelCase__ : List[str]=True , lowerCamelCase__ : List[str]=None , **lowerCamelCase__ : Optional[Any] , ):
a__ : Optional[int] = self._strip_protocol(lowerCamelCase__ )
if mode != "rb":
raise ValueError(f'''Tried to read with mode {mode} on file {self.file.path} opened with mode \'rb\'''' )
return self.file.open()
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'bz2'
_lowercase = 'bz2'
_lowercase = '.bz2'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gzip'
_lowercase = 'gzip'
_lowercase = '.gz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'lz4'
_lowercase = 'lz4'
_lowercase = '.lz4'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'xz'
_lowercase = 'xz'
_lowercase = '.xz'
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'zstd'
_lowercase = 'zstd'
_lowercase = '.zst'
def __init__( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : str = "rb" , lowerCamelCase__ : Optional[str] = None , lowerCamelCase__ : Optional[dict] = None , lowerCamelCase__ : int = DEFAULT_BLOCK_SIZE , **lowerCamelCase__ : Tuple , ):
super().__init__(
fo=lowerCamelCase__ , mode=lowerCamelCase__ , target_protocol=lowerCamelCase__ , target_options=lowerCamelCase__ , block_size=lowerCamelCase__ , **lowerCamelCase__ , )
# We need to wrap the zstd decompressor to avoid this error in fsspec==2021.7.0 and zstandard==0.15.2:
#
# File "/Users/user/.virtualenvs/hf-datasets/lib/python3.7/site-packages/fsspec/core.py", line 145, in open
# out.close = close
# AttributeError: 'zstd.ZstdDecompressionReader' object attribute 'close' is read-only
#
# see https://github.com/intake/filesystem_spec/issues/725
a__ : Any = self.file.__enter__
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : str ):
a__ : List[Any] = file_
def __enter__( self : str ):
self._file.__enter__()
return self
def __exit__( self : int , *lowerCamelCase__ : List[str] , **lowerCamelCase__ : int ):
self._file.__exit__(*lowerCamelCase__ , **lowerCamelCase__ )
def __iter__( self : List[str] ):
return iter(self._file )
def _UpperCamelCase( self : Any ):
return next(self._file )
def __getattr__( self : Optional[Any] , lowerCamelCase__ : Tuple ):
return getattr(self._file , lowerCamelCase__ )
def fixed_enter(*lowerCamelCase__ : List[str] , **lowerCamelCase__ : str ):
return WrappedFile(_enter(*lowerCamelCase__ , **lowerCamelCase__ ) )
a__ : Any = fixed_enter
| 37 |
import json
from typing import Dict, List, Optional, Tuple, Union
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding, EncodedInput
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import PaddingStrategy, logging
from .tokenization_led import LEDTokenizer
UpperCamelCase : Union[str, Any] = logging.get_logger(__name__)
UpperCamelCase : Optional[Any] = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Optional[Any] = {
"""vocab_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json""",
},
"""merges_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt""",
},
"""tokenizer_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json""",
},
}
UpperCamelCase : Dict = {
"""allenai/led-base-16384""": 1_6384,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = LEDTokenizer
_lowercase = ['input_ids', 'attention_mask']
def __init__( self : Tuple , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[str]=None , lowerCamelCase__ : Any=None , lowerCamelCase__ : int="replace" , lowerCamelCase__ : Union[str, Any]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Tuple="</s>" , lowerCamelCase__ : Optional[int]="<s>" , lowerCamelCase__ : str="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Any="<mask>" , lowerCamelCase__ : Optional[int]=False , lowerCamelCase__ : int=True , **lowerCamelCase__ : Union[str, Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : List[str] = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : Optional[Any] = add_prefix_space
a__ : List[str] = pre_tok_class(**lowerCamelCase__ )
a__ : Optional[int] = add_prefix_space
# the pre_tokenizer is already updated in the GPT2TokenizerFast `__init__`
a__ : Any = "post_processor"
a__ : str = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Any = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Optional[Any] = tuple(state["sep"] )
if "cls" in state:
a__ : Optional[Any] = tuple(state["cls"] )
a__ : Optional[int] = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Dict = add_prefix_space
a__ : int = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : List[Any] = trim_offsets
a__ : List[str] = True
if changes_to_apply:
a__ : int = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : int = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
# Copied from transformers.models.bart.tokenization_bart_fast.BartTokenizerFast.mask_token with BART->LED
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Union[str, Any] ):
a__ : Any = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : Union[str, Any] = value
def _UpperCamelCase( self : Any , *lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : List[str] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
if is_split_into_words and not self.add_prefix_space:
raise ValueError(
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs." )
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Any , *lowerCamelCase__ : Dict , **lowerCamelCase__ : Optional[Any] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
if is_split_into_words and not self.add_prefix_space:
raise ValueError(
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs." )
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : List[str] = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[Any]=None ):
a__ : Any = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : List[str] = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Union[Dict[str, EncodedInput], BatchEncoding] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : PaddingStrategy = PaddingStrategy.DO_NOT_PAD , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[bool] = None , ):
a__ : str = super()._pad(
encoded_inputs=lowerCamelCase__ , max_length=lowerCamelCase__ , padding_strategy=lowerCamelCase__ , pad_to_multiple_of=lowerCamelCase__ , return_attention_mask=lowerCamelCase__ , )
# Load from model defaults
if return_attention_mask is None:
a__ : Optional[int] = "attention_mask" in self.model_input_names
if return_attention_mask and "global_attention_mask" in encoded_inputs:
a__ : Tuple = encoded_inputs[self.model_input_names[0]]
# `global_attention_mask` need to have the same length as other (sequential) inputs.
a__ : Dict = len(encoded_inputs["global_attention_mask"] ) != len(lowerCamelCase__ )
if needs_to_be_padded:
a__ : Union[str, Any] = len(lowerCamelCase__ ) - len(encoded_inputs["global_attention_mask"] )
if self.padding_side == "right":
# Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend`
a__ : List[Any] = (
encoded_inputs["global_attention_mask"] + [-1] * difference
)
elif self.padding_side == "left":
a__ : Any = [-1] * difference + encoded_inputs[
"global_attention_mask"
]
else:
raise ValueError("Invalid padding strategy:" + str(self.padding_side ) )
return encoded_inputs
| 37 | 1 |
import argparse
import json
import os
from collections import OrderedDict
import torch
from transformers import LukeConfig, LukeForMaskedLM, MLukeTokenizer, XLMRobertaTokenizer
from transformers.tokenization_utils_base import AddedToken
@torch.no_grad()
def UpperCamelCase_ ( __a , __a , __a , __a , __a ) -> Tuple:
# Load configuration defined in the metadata file
with open(__a ) as metadata_file:
a__ : Optional[int] = json.load(__a )
a__ : str = LukeConfig(use_entity_aware_attention=__a , **metadata["model_config"] )
# Load in the weights from the checkpoint_path
a__ : List[Any] = torch.load(__a , map_location="cpu" )["module"]
# Load the entity vocab file
a__ : Union[str, Any] = load_original_entity_vocab(__a )
# add an entry for [MASK2]
a__ : List[Any] = max(entity_vocab.values() ) + 1
config.entity_vocab_size += 1
a__ : Optional[int] = XLMRobertaTokenizer.from_pretrained(metadata["model_config"]["bert_model_name"] )
# Add special tokens to the token vocabulary for downstream tasks
a__ : Union[str, Any] = AddedToken("<ent>" , lstrip=__a , rstrip=__a )
a__ : Dict = AddedToken("<ent2>" , lstrip=__a , rstrip=__a )
tokenizer.add_special_tokens({"additional_special_tokens": [entity_token_a, entity_token_a]} )
config.vocab_size += 2
print(f'''Saving tokenizer to {pytorch_dump_folder_path}''' )
tokenizer.save_pretrained(__a )
with open(os.path.join(__a , "tokenizer_config.json" ) , "r" ) as f:
a__ : int = json.load(__a )
a__ : List[str] = "MLukeTokenizer"
with open(os.path.join(__a , "tokenizer_config.json" ) , "w" ) as f:
json.dump(__a , __a )
with open(os.path.join(__a , MLukeTokenizer.vocab_files_names["entity_vocab_file"] ) , "w" ) as f:
json.dump(__a , __a )
a__ : Dict = MLukeTokenizer.from_pretrained(__a )
# Initialize the embeddings of the special tokens
a__ : Dict = tokenizer.convert_tokens_to_ids(["@"] )[0]
a__ : Union[str, Any] = tokenizer.convert_tokens_to_ids(["#"] )[0]
a__ : List[str] = state_dict["embeddings.word_embeddings.weight"]
a__ : Optional[int] = word_emb[ent_init_index].unsqueeze(0 )
a__ : Union[str, Any] = word_emb[enta_init_index].unsqueeze(0 )
a__ : List[Any] = torch.cat([word_emb, ent_emb, enta_emb] )
# add special tokens for 'entity_predictions.bias'
for bias_name in ["lm_head.decoder.bias", "lm_head.bias"]:
a__ : str = state_dict[bias_name]
a__ : Optional[int] = decoder_bias[ent_init_index].unsqueeze(0 )
a__ : str = decoder_bias[enta_init_index].unsqueeze(0 )
a__ : Tuple = torch.cat([decoder_bias, ent_decoder_bias, enta_decoder_bias] )
# Initialize the query layers of the entity-aware self-attention mechanism
for layer_index in range(config.num_hidden_layers ):
for matrix_name in ["query.weight", "query.bias"]:
a__ : Optional[Any] = f'''encoder.layer.{layer_index}.attention.self.'''
a__ : List[str] = state_dict[prefix + matrix_name]
a__ : str = state_dict[prefix + matrix_name]
a__ : Any = state_dict[prefix + matrix_name]
# Initialize the embedding of the [MASK2] entity using that of the [MASK] entity for downstream tasks
a__ : List[Any] = state_dict["entity_embeddings.entity_embeddings.weight"]
a__ : Union[str, Any] = entity_emb[entity_vocab["[MASK]"]].unsqueeze(0 )
a__ : List[Any] = torch.cat([entity_emb, entity_mask_emb] )
# add [MASK2] for 'entity_predictions.bias'
a__ : List[str] = state_dict["entity_predictions.bias"]
a__ : Dict = entity_prediction_bias[entity_vocab["[MASK]"]].unsqueeze(0 )
a__ : Any = torch.cat([entity_prediction_bias, entity_mask_bias] )
a__ : int = LukeForMaskedLM(config=__a ).eval()
state_dict.pop("entity_predictions.decoder.weight" )
state_dict.pop("lm_head.decoder.weight" )
state_dict.pop("lm_head.decoder.bias" )
a__ : Dict = OrderedDict()
for key, value in state_dict.items():
if not (key.startswith("lm_head" ) or key.startswith("entity_predictions" )):
a__ : Union[str, Any] = state_dict[key]
else:
a__ : Dict = state_dict[key]
a__, a__ : List[str] = model.load_state_dict(__a , strict=__a )
if set(__a ) != {"luke.embeddings.position_ids"}:
raise ValueError(f'''Unexpected unexpected_keys: {unexpected_keys}''' )
if set(__a ) != {
"lm_head.decoder.weight",
"lm_head.decoder.bias",
"entity_predictions.decoder.weight",
}:
raise ValueError(f'''Unexpected missing_keys: {missing_keys}''' )
model.tie_weights()
assert (model.luke.embeddings.word_embeddings.weight == model.lm_head.decoder.weight).all()
assert (model.luke.entity_embeddings.entity_embeddings.weight == model.entity_predictions.decoder.weight).all()
# Check outputs
a__ : int = MLukeTokenizer.from_pretrained(__a , task="entity_classification" )
a__ : int = "ISO 639-3 uses the code fas for the dialects spoken across Iran and アフガニスタン (Afghanistan)."
a__ : Optional[Any] = (0, 9)
a__ : int = tokenizer(__a , entity_spans=[span] , return_tensors="pt" )
a__ : Optional[int] = model(**__a )
# Verify word hidden states
if model_size == "large":
raise NotImplementedError
else: # base
a__ : List[Any] = torch.Size((1, 33, 768) )
a__ : int = torch.tensor([[0.0892, 0.0596, -0.2819], [0.0134, 0.1199, 0.0573], [-0.0169, 0.0927, 0.0644]] )
if not (outputs.last_hidden_state.shape == expected_shape):
raise ValueError(
f'''Outputs.last_hidden_state.shape is {outputs.last_hidden_state.shape}, Expected shape is {expected_shape}''' )
if not torch.allclose(outputs.last_hidden_state[0, :3, :3] , __a , atol=1e-4 ):
raise ValueError
# Verify entity hidden states
if model_size == "large":
raise NotImplementedError
else: # base
a__ : str = torch.Size((1, 1, 768) )
a__ : Any = torch.tensor([[-0.1482, 0.0609, 0.0322]] )
if not (outputs.entity_last_hidden_state.shape == expected_shape):
raise ValueError(
f'''Outputs.entity_last_hidden_state.shape is {outputs.entity_last_hidden_state.shape}, Expected shape is'''
f''' {expected_shape}''' )
if not torch.allclose(outputs.entity_last_hidden_state[0, :3, :3] , __a , atol=1e-4 ):
raise ValueError
# Verify masked word/entity prediction
a__ : int = MLukeTokenizer.from_pretrained(__a )
a__ : int = "Tokyo is the capital of <mask>."
a__ : Any = (24, 30)
a__ : Optional[int] = tokenizer(__a , entity_spans=[span] , return_tensors="pt" )
a__ : Union[str, Any] = model(**__a )
a__ : Any = encoding["input_ids"][0].tolist()
a__ : int = input_ids.index(tokenizer.convert_tokens_to_ids("<mask>" ) )
a__ : Any = outputs.logits[0][mask_position_id].argmax(dim=-1 )
assert "Japan" == tokenizer.decode(__a )
a__ : List[Any] = outputs.entity_logits[0][0].argmax().item()
a__ : Tuple = [
entity for entity, entity_id in tokenizer.entity_vocab.items() if entity_id == predicted_entity_id
]
assert [e for e in multilingual_predicted_entities if e.startswith("en:" )][0] == "en:Japan"
# Finally, save our PyTorch model and tokenizer
print("Saving PyTorch model to {}".format(__a ) )
model.save_pretrained(__a )
def UpperCamelCase_ ( __a ) -> int:
a__ : str = ["[MASK]", "[PAD]", "[UNK]"]
a__ : List[Any] = [json.loads(__a ) for line in open(__a )]
a__ : Optional[Any] = {}
for entry in data:
a__ : Optional[int] = entry["id"]
for entity_name, language in entry["entities"]:
if entity_name in SPECIAL_TOKENS:
a__ : Dict = entity_id
break
a__ : Union[str, Any] = f'''{language}:{entity_name}'''
a__ : Union[str, Any] = entity_id
return new_mapping
if __name__ == "__main__":
UpperCamelCase : Dict = argparse.ArgumentParser()
# Required parameters
parser.add_argument("""--checkpoint_path""", type=str, help="""Path to a pytorch_model.bin file.""")
parser.add_argument(
"""--metadata_path""", default=None, type=str, help="""Path to a metadata.json file, defining the configuration."""
)
parser.add_argument(
"""--entity_vocab_path""",
default=None,
type=str,
help="""Path to an entity_vocab.tsv file, containing the entity vocabulary.""",
)
parser.add_argument(
"""--pytorch_dump_folder_path""", default=None, type=str, help="""Path to where to dump the output PyTorch model."""
)
parser.add_argument(
"""--model_size""", default="""base""", type=str, choices=["""base""", """large"""], help="""Size of the model to be converted."""
)
UpperCamelCase : Tuple = parser.parse_args()
convert_luke_checkpoint(
args.checkpoint_path,
args.metadata_path,
args.entity_vocab_path,
args.pytorch_dump_folder_path,
args.model_size,
)
| 37 |
import json
from typing import List, Optional, Tuple
from tokenizers import pre_tokenizers, processors
from ...tokenization_utils_base import AddedToken, BatchEncoding
from ...tokenization_utils_fast import PreTrainedTokenizerFast
from ...utils import logging
from .tokenization_roberta import RobertaTokenizer
UpperCamelCase : Any = logging.get_logger(__name__)
UpperCamelCase : Any = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt""", """tokenizer_file""": """tokenizer.json"""}
UpperCamelCase : Union[str, Any] = {
"""vocab_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/vocab.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/vocab.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/vocab.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/vocab.json""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/vocab.json""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/vocab.json"""
),
},
"""merges_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/merges.txt""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/merges.txt""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/merges.txt""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/merges.txt""",
"""roberta-base-openai-detector""": """https://huggingface.co/roberta-base-openai-detector/resolve/main/merges.txt""",
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/merges.txt"""
),
},
"""tokenizer_file""": {
"""roberta-base""": """https://huggingface.co/roberta-base/resolve/main/tokenizer.json""",
"""roberta-large""": """https://huggingface.co/roberta-large/resolve/main/tokenizer.json""",
"""roberta-large-mnli""": """https://huggingface.co/roberta-large-mnli/resolve/main/tokenizer.json""",
"""distilroberta-base""": """https://huggingface.co/distilroberta-base/resolve/main/tokenizer.json""",
"""roberta-base-openai-detector""": (
"""https://huggingface.co/roberta-base-openai-detector/resolve/main/tokenizer.json"""
),
"""roberta-large-openai-detector""": (
"""https://huggingface.co/roberta-large-openai-detector/resolve/main/tokenizer.json"""
),
},
}
UpperCamelCase : List[str] = {
"""roberta-base""": 512,
"""roberta-large""": 512,
"""roberta-large-mnli""": 512,
"""distilroberta-base""": 512,
"""roberta-base-openai-detector""": 512,
"""roberta-large-openai-detector""": 512,
}
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = ['input_ids', 'attention_mask']
_lowercase = RobertaTokenizer
def __init__( self : List[str] , lowerCamelCase__ : Any=None , lowerCamelCase__ : List[Any]=None , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]="replace" , lowerCamelCase__ : List[str]="<s>" , lowerCamelCase__ : Union[str, Any]="</s>" , lowerCamelCase__ : Any="</s>" , lowerCamelCase__ : Any="<s>" , lowerCamelCase__ : int="<unk>" , lowerCamelCase__ : Any="<pad>" , lowerCamelCase__ : Tuple="<mask>" , lowerCamelCase__ : Any=False , lowerCamelCase__ : Dict=True , **lowerCamelCase__ : Optional[Any] , ):
super().__init__(
lowerCamelCase__ , lowerCamelCase__ , tokenizer_file=lowerCamelCase__ , errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , trim_offsets=lowerCamelCase__ , **lowerCamelCase__ , )
a__ : List[Any] = json.loads(self.backend_tokenizer.pre_tokenizer.__getstate__() )
if pre_tok_state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : Any = getattr(lowerCamelCase__ , pre_tok_state.pop("type" ) )
a__ : int = add_prefix_space
a__ : Tuple = pre_tok_class(**lowerCamelCase__ )
a__ : str = add_prefix_space
a__ : Tuple = "post_processor"
a__ : Dict = getattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
if tokenizer_component_instance:
a__ : Tuple = json.loads(tokenizer_component_instance.__getstate__() )
# The lists 'sep' and 'cls' must be cased in tuples for the object `post_processor_class`
if "sep" in state:
a__ : Tuple = tuple(state["sep"] )
if "cls" in state:
a__ : str = tuple(state["cls"] )
a__ : str = False
if state.get("add_prefix_space" , lowerCamelCase__ ) != add_prefix_space:
a__ : str = add_prefix_space
a__ : Any = True
if state.get("trim_offsets" , lowerCamelCase__ ) != trim_offsets:
a__ : int = trim_offsets
a__ : Dict = True
if changes_to_apply:
a__ : Union[str, Any] = getattr(lowerCamelCase__ , state.pop("type" ) )
a__ : str = component_class(**lowerCamelCase__ )
setattr(self.backend_tokenizer , lowerCamelCase__ , lowerCamelCase__ )
@property
def _UpperCamelCase( self : Union[str, Any] ):
if self._mask_token is None:
if self.verbose:
logger.error("Using mask_token, but it is not set yet." )
return None
return str(self._mask_token )
@mask_token.setter
def _UpperCamelCase( self : List[Any] , lowerCamelCase__ : Tuple ):
a__ : List[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else value
a__ : List[str] = value
def _UpperCamelCase( self : Union[str, Any] , *lowerCamelCase__ : int , **lowerCamelCase__ : int ):
a__ : Optional[int] = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._batch_encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : Tuple , *lowerCamelCase__ : Dict , **lowerCamelCase__ : List[str] ):
a__ : Dict = kwargs.get("is_split_into_words" , lowerCamelCase__ )
assert self.add_prefix_space or not is_split_into_words, (
f'''You need to instantiate {self.__class__.__name__} with add_prefix_space=True '''
"to use it with pretokenized inputs."
)
return super()._encode_plus(*lowerCamelCase__ , **lowerCamelCase__ )
def _UpperCamelCase( self : str , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
a__ : int = self._tokenizer.model.save(lowerCamelCase__ , name=lowerCamelCase__ )
return tuple(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Dict , lowerCamelCase__ : Optional[int]=None ):
a__ : Union[str, Any] = [self.bos_token_id] + token_ids_a + [self.eos_token_id]
if token_ids_a is None:
return output
return output + [self.eos_token_id] + token_ids_a + [self.eos_token_id]
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : Tuple = [self.sep_token_id]
a__ : int = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
| 37 | 1 |
import argparse
import os
from . import (
ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
BART_PRETRAINED_MODEL_ARCHIVE_LIST,
BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP,
DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
DPR_CONTEXT_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
DPR_QUESTION_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
DPR_READER_PRETRAINED_MODEL_ARCHIVE_LIST,
ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP,
FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP,
LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST,
LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP,
ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
T5_PRETRAINED_CONFIG_ARCHIVE_MAP,
TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP,
WAV_2_VEC_2_PRETRAINED_CONFIG_ARCHIVE_MAP,
XLM_PRETRAINED_CONFIG_ARCHIVE_MAP,
XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
AlbertConfig,
BartConfig,
BertConfig,
CamembertConfig,
CTRLConfig,
DistilBertConfig,
DPRConfig,
ElectraConfig,
FlaubertConfig,
GPTaConfig,
LayoutLMConfig,
LxmertConfig,
OpenAIGPTConfig,
RobertaConfig,
TaConfig,
TFAlbertForPreTraining,
TFBartForConditionalGeneration,
TFBartForSequenceClassification,
TFBertForPreTraining,
TFBertForQuestionAnswering,
TFBertForSequenceClassification,
TFCamembertForMaskedLM,
TFCTRLLMHeadModel,
TFDistilBertForMaskedLM,
TFDistilBertForQuestionAnswering,
TFDPRContextEncoder,
TFDPRQuestionEncoder,
TFDPRReader,
TFElectraForPreTraining,
TFFlaubertWithLMHeadModel,
TFGPTaLMHeadModel,
TFLayoutLMForMaskedLM,
TFLxmertForPreTraining,
TFLxmertVisualFeatureEncoder,
TFOpenAIGPTLMHeadModel,
TFRobertaForCausalLM,
TFRobertaForMaskedLM,
TFRobertaForSequenceClassification,
TFTaForConditionalGeneration,
TFTransfoXLLMHeadModel,
TFWavaVecaModel,
TFXLMRobertaForMaskedLM,
TFXLMWithLMHeadModel,
TFXLNetLMHeadModel,
TransfoXLConfig,
WavaVecaConfig,
WavaVecaModel,
XLMConfig,
XLMRobertaConfig,
XLNetConfig,
is_torch_available,
load_pytorch_checkpoint_in_tfa_model,
)
from .utils import CONFIG_NAME, WEIGHTS_NAME, cached_file, logging
if is_torch_available():
import numpy as np
import torch
from . import (
AlbertForPreTraining,
BartForConditionalGeneration,
BertForPreTraining,
BertForQuestionAnswering,
BertForSequenceClassification,
CamembertForMaskedLM,
CTRLLMHeadModel,
DistilBertForMaskedLM,
DistilBertForQuestionAnswering,
DPRContextEncoder,
DPRQuestionEncoder,
DPRReader,
ElectraForPreTraining,
FlaubertWithLMHeadModel,
GPTaLMHeadModel,
LayoutLMForMaskedLM,
LxmertForPreTraining,
LxmertVisualFeatureEncoder,
OpenAIGPTLMHeadModel,
RobertaForMaskedLM,
RobertaForSequenceClassification,
TaForConditionalGeneration,
TransfoXLLMHeadModel,
XLMRobertaForMaskedLM,
XLMWithLMHeadModel,
XLNetLMHeadModel,
)
logging.set_verbosity_info()
UpperCamelCase : Dict = {
"""bart""": (
BartConfig,
TFBartForConditionalGeneration,
TFBartForSequenceClassification,
BartForConditionalGeneration,
BART_PRETRAINED_MODEL_ARCHIVE_LIST,
),
"""bert""": (
BertConfig,
TFBertForPreTraining,
BertForPreTraining,
BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""bert-large-uncased-whole-word-masking-finetuned-squad""": (
BertConfig,
TFBertForQuestionAnswering,
BertForQuestionAnswering,
BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""bert-large-cased-whole-word-masking-finetuned-squad""": (
BertConfig,
TFBertForQuestionAnswering,
BertForQuestionAnswering,
BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""bert-base-cased-finetuned-mrpc""": (
BertConfig,
TFBertForSequenceClassification,
BertForSequenceClassification,
BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""dpr""": (
DPRConfig,
TFDPRQuestionEncoder,
TFDPRContextEncoder,
TFDPRReader,
DPRQuestionEncoder,
DPRContextEncoder,
DPRReader,
DPR_CONTEXT_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
DPR_QUESTION_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
DPR_READER_PRETRAINED_MODEL_ARCHIVE_LIST,
),
"""gpt2""": (
GPTaConfig,
TFGPTaLMHeadModel,
GPTaLMHeadModel,
GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""xlnet""": (
XLNetConfig,
TFXLNetLMHeadModel,
XLNetLMHeadModel,
XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""xlm""": (
XLMConfig,
TFXLMWithLMHeadModel,
XLMWithLMHeadModel,
XLM_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""xlm-roberta""": (
XLMRobertaConfig,
TFXLMRobertaForMaskedLM,
XLMRobertaForMaskedLM,
XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""transfo-xl""": (
TransfoXLConfig,
TFTransfoXLLMHeadModel,
TransfoXLLMHeadModel,
TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""openai-gpt""": (
OpenAIGPTConfig,
TFOpenAIGPTLMHeadModel,
OpenAIGPTLMHeadModel,
OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""roberta""": (
RobertaConfig,
TFRobertaForCausalLM,
TFRobertaForMaskedLM,
RobertaForMaskedLM,
ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""layoutlm""": (
LayoutLMConfig,
TFLayoutLMForMaskedLM,
LayoutLMForMaskedLM,
LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST,
),
"""roberta-large-mnli""": (
RobertaConfig,
TFRobertaForSequenceClassification,
RobertaForSequenceClassification,
ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""camembert""": (
CamembertConfig,
TFCamembertForMaskedLM,
CamembertForMaskedLM,
CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""flaubert""": (
FlaubertConfig,
TFFlaubertWithLMHeadModel,
FlaubertWithLMHeadModel,
FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""distilbert""": (
DistilBertConfig,
TFDistilBertForMaskedLM,
DistilBertForMaskedLM,
DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""distilbert-base-distilled-squad""": (
DistilBertConfig,
TFDistilBertForQuestionAnswering,
DistilBertForQuestionAnswering,
DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""lxmert""": (
LxmertConfig,
TFLxmertForPreTraining,
LxmertForPreTraining,
LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""lxmert-visual-feature-encoder""": (
LxmertConfig,
TFLxmertVisualFeatureEncoder,
LxmertVisualFeatureEncoder,
LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""ctrl""": (
CTRLConfig,
TFCTRLLMHeadModel,
CTRLLMHeadModel,
CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""albert""": (
AlbertConfig,
TFAlbertForPreTraining,
AlbertForPreTraining,
ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""t5""": (
TaConfig,
TFTaForConditionalGeneration,
TaForConditionalGeneration,
T5_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""electra""": (
ElectraConfig,
TFElectraForPreTraining,
ElectraForPreTraining,
ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
"""wav2vec2""": (
WavaVecaConfig,
TFWavaVecaModel,
WavaVecaModel,
WAV_2_VEC_2_PRETRAINED_CONFIG_ARCHIVE_MAP,
),
}
def UpperCamelCase_ ( __a , __a , __a , __a , __a=False , __a=True ) -> Tuple:
if model_type not in MODEL_CLASSES:
raise ValueError(f'''Unrecognized model type, should be one of {list(MODEL_CLASSES.keys() )}.''' )
a__, a__, a__, a__ : List[str] = MODEL_CLASSES[model_type]
# Initialise TF model
if config_file in aws_config_map:
a__ : Optional[Any] = cached_file(__a , __a , force_download=not use_cached_models )
a__ : Tuple = config_class.from_json_file(__a )
a__ : List[Any] = True
a__ : Optional[int] = True
print(f'''Building TensorFlow model from configuration: {config}''' )
a__ : List[str] = model_class(__a )
# Load weights from tf checkpoint
if pytorch_checkpoint_path in aws_config_map.keys():
a__ : int = cached_file(
__a , __a , force_download=not use_cached_models )
# Load PyTorch checkpoint in tf2 model:
a__ : List[str] = load_pytorch_checkpoint_in_tfa_model(__a , __a )
if compare_with_pt_model:
a__ : Optional[int] = tf_model(tf_model.dummy_inputs , training=__a ) # build the network
a__ : Union[str, Any] = torch.load(__a , map_location="cpu" )
a__ : Optional[Any] = pt_model_class.from_pretrained(
pretrained_model_name_or_path=__a , config=__a , state_dict=__a )
with torch.no_grad():
a__ : Dict = pt_model(**pt_model.dummy_inputs )
a__ : str = pto[0].numpy()
a__ : Union[str, Any] = tfo[0].numpy()
a__ : Any = np.amax(np.abs(np_pt - np_tf ) )
print(f'''Max absolute difference between models outputs {diff}''' )
assert diff <= 2e-2, f'''Error, model absolute difference is >2e-2: {diff}'''
# Save pytorch-model
print(f'''Save TensorFlow model to {tf_dump_path}''' )
tf_model.save_weights(__a , save_format="h5" )
def UpperCamelCase_ ( __a , __a , __a=None , __a=None , __a=False , __a=False , __a=False , __a=False , ) -> Dict:
if args_model_type is None:
a__ : str = list(MODEL_CLASSES.keys() )
else:
a__ : int = [args_model_type]
for j, model_type in enumerate(__a , start=1 ):
print("=" * 100 )
print(f''' Converting model type {j}/{len(__a )}: {model_type}''' )
print("=" * 100 )
if model_type not in MODEL_CLASSES:
raise ValueError(f'''Unrecognized model type {model_type}, should be one of {list(MODEL_CLASSES.keys() )}.''' )
a__, a__, a__, a__, a__ : List[Any] = MODEL_CLASSES[model_type]
if model_shortcut_names_or_path is None:
a__ : int = list(aws_model_maps.keys() )
if config_shortcut_names_or_path is None:
a__ : Optional[Any] = model_shortcut_names_or_path
for i, (model_shortcut_name, config_shortcut_name) in enumerate(
zip(__a , __a ) , start=1 ):
print("-" * 100 )
if "-squad" in model_shortcut_name or "-mrpc" in model_shortcut_name or "-mnli" in model_shortcut_name:
if not only_convert_finetuned_models:
print(f''' Skipping finetuned checkpoint {model_shortcut_name}''' )
continue
a__ : Any = model_shortcut_name
elif only_convert_finetuned_models:
print(f''' Skipping not finetuned checkpoint {model_shortcut_name}''' )
continue
print(
f''' Converting checkpoint {i}/{len(__a )}: {model_shortcut_name} - model_type {model_type}''' )
print("-" * 100 )
if config_shortcut_name in aws_config_map:
a__ : Dict = cached_file(__a , __a , force_download=not use_cached_models )
else:
a__ : str = config_shortcut_name
if model_shortcut_name in aws_model_maps:
a__ : List[str] = cached_file(__a , __a , force_download=not use_cached_models )
else:
a__ : Optional[Any] = model_shortcut_name
if os.path.isfile(__a ):
a__ : int = "converted_model"
convert_pt_checkpoint_to_tf(
model_type=__a , pytorch_checkpoint_path=__a , config_file=__a , tf_dump_path=os.path.join(__a , model_shortcut_name + "-tf_model.h5" ) , compare_with_pt_model=__a , )
if remove_cached_files:
os.remove(__a )
os.remove(__a )
if __name__ == "__main__":
UpperCamelCase : List[str] = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"""--tf_dump_path""", default=None, type=str, required=True, help="""Path to the output Tensorflow dump file."""
)
parser.add_argument(
"""--model_type""",
default=None,
type=str,
help=(
f"""Model type selected in the list of {list(MODEL_CLASSES.keys())}. If not given, will download and """
"""convert all the models from AWS."""
),
)
parser.add_argument(
"""--pytorch_checkpoint_path""",
default=None,
type=str,
help=(
"""Path to the PyTorch checkpoint path or shortcut name to download from AWS. """
"""If not given, will download and convert all the checkpoints from AWS."""
),
)
parser.add_argument(
"""--config_file""",
default=None,
type=str,
help=(
"""The config json file corresponding to the pre-trained model. \n"""
"""This specifies the model architecture. If not given and """
"""--pytorch_checkpoint_path is not given or is a shortcut name """
"""use the configuration associated to the shortcut name on the AWS"""
),
)
parser.add_argument(
"""--compare_with_pt_model""", action="""store_true""", help="""Compare Tensorflow and PyTorch model predictions."""
)
parser.add_argument(
"""--use_cached_models""",
action="""store_true""",
help="""Use cached models if possible instead of updating to latest checkpoint versions.""",
)
parser.add_argument(
"""--remove_cached_files""",
action="""store_true""",
help="""Remove pytorch models after conversion (save memory when converting in batches).""",
)
parser.add_argument("""--only_convert_finetuned_models""", action="""store_true""", help="""Only convert finetuned models.""")
UpperCamelCase : List[Any] = parser.parse_args()
# if args.pytorch_checkpoint_path is not None:
# convert_pt_checkpoint_to_tf(args.model_type.lower(),
# args.pytorch_checkpoint_path,
# args.config_file if args.config_file is not None else args.pytorch_checkpoint_path,
# args.tf_dump_path,
# compare_with_pt_model=args.compare_with_pt_model,
# use_cached_models=args.use_cached_models)
# else:
convert_all_pt_checkpoints_to_tf(
args.model_type.lower() if args.model_type is not None else None,
args.tf_dump_path,
model_shortcut_names_or_path=[args.pytorch_checkpoint_path]
if args.pytorch_checkpoint_path is not None
else None,
config_shortcut_names_or_path=[args.config_file] if args.config_file is not None else None,
compare_with_pt_model=args.compare_with_pt_model,
use_cached_models=args.use_cached_models,
remove_cached_files=args.remove_cached_files,
only_convert_finetuned_models=args.only_convert_finetuned_models,
)
| 37 |
from statistics import mean, stdev
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : List[str] = min(__a )
a__ : str = max(__a )
# normalize data
return [round((x - x_min) / (x_max - x_min) , __a ) for x in data]
def UpperCamelCase_ ( __a , __a = 3 ) -> list:
a__ : str = mean(__a )
a__ : List[str] = stdev(__a )
# standardize data
return [round((x - mu) / (sigma) , __a ) for x in data]
| 37 | 1 |
from ....utils import logging
UpperCamelCase : Optional[int] = logging.get_logger(__name__)
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Tuple , lowerCamelCase__ : Any , lowerCamelCase__ : List[str]=None , lowerCamelCase__ : Tuple=2_048 ):
a__ : Optional[Any] = config.__dict__
a__ : List[Any] = modal_hidden_size
if num_labels:
a__ : Tuple = num_labels
| 37 |
def UpperCamelCase_ ( __a = 50 ) -> int:
a__ : Tuple = [[0] * 3 for _ in range(length + 1 )]
for row_length in range(length + 1 ):
for tile_length in range(2 , 5 ):
for tile_start in range(row_length - tile_length + 1 ):
different_colour_ways_number[row_length][tile_length - 2] += (
different_colour_ways_number[row_length - tile_start - tile_length][
tile_length - 2
]
+ 1
)
return sum(different_colour_ways_number[length] )
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 | 1 |
import json
import logging
import os
import sys
from pathlib import Path
import finetune_rag
from transformers.file_utils import is_apex_available
from transformers.testing_utils import (
TestCasePlus,
execute_subprocess_async,
require_ray,
require_torch_gpu,
require_torch_multi_gpu,
)
logging.basicConfig(level=logging.DEBUG)
UpperCamelCase : Optional[int] = logging.getLogger()
UpperCamelCase : Union[str, Any] = logging.StreamHandler(sys.stdout)
logger.addHandler(stream_handler)
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Tuple ):
os.makedirs(lowerCamelCase__ , exist_ok=lowerCamelCase__ )
a__ : Optional[Any] = {"source": "What is love ?", "target": "life"}
a__ : int = {"train": 12, "val": 2, "test": 2}
for split in ["train", "test", "val"]:
for field in ["source", "target"]:
a__ : Any = "\n".join([contents[field]] * n_lines[split] )
with open(os.path.join(lowerCamelCase__ , f'''{split}.{field}''' ) , "w" ) as f:
f.write(lowerCamelCase__ )
def _UpperCamelCase( self : int , lowerCamelCase__ : int , lowerCamelCase__ : str = "pytorch" ):
a__ : List[Any] = self.get_auto_remove_tmp_dir()
a__ : Dict = os.path.join(lowerCamelCase__ , "output" )
a__ : List[Any] = os.path.join(lowerCamelCase__ , "data" )
self._create_dummy_data(data_dir=lowerCamelCase__ )
a__ : Optional[Any] = f'''
--data_dir {data_dir} \
--output_dir {output_dir} \
--model_name_or_path facebook/rag-sequence-base \
--model_type rag_sequence \
--do_train \
--do_predict \
--n_val -1 \
--val_check_interval 1.0 \
--train_batch_size 2 \
--eval_batch_size 1 \
--max_source_length 25 \
--max_target_length 25 \
--val_max_target_length 25 \
--test_max_target_length 25 \
--label_smoothing 0.1 \
--dropout 0.1 \
--attention_dropout 0.1 \
--weight_decay 0.001 \
--adam_epsilon 1e-08 \
--max_grad_norm 0.1 \
--lr_scheduler polynomial \
--learning_rate 3e-04 \
--num_train_epochs 1 \
--warmup_steps 4 \
--gradient_accumulation_steps 1 \
--distributed-port 8787 \
--use_dummy_dataset 1 \
--distributed_retriever {distributed_retriever} \
'''.split()
if gpus > 0:
testargs.append(f'''--gpus={gpus}''' )
if is_apex_available():
testargs.append("--fp16" )
else:
testargs.append("--gpus=0" )
testargs.append("--distributed_backend=ddp_cpu" )
testargs.append("--num_processes=2" )
a__ : Optional[int] = [sys.executable, str(Path(finetune_rag.__file__ ).resolve() )] + testargs
execute_subprocess_async(lowerCamelCase__ , env=self.get_env() )
a__ : List[Any] = os.path.join(lowerCamelCase__ , "metrics.json" )
with open(lowerCamelCase__ ) as f:
a__ : str = json.load(lowerCamelCase__ )
return result
@require_torch_gpu
def _UpperCamelCase( self : List[str] ):
a__ : List[Any] = self._run_finetune(gpus=1 )
self.assertGreaterEqual(result["test"][0]["test_avg_em"] , 0.2 )
@require_torch_multi_gpu
def _UpperCamelCase( self : Optional[Any] ):
a__ : Union[str, Any] = self._run_finetune(gpus=2 )
self.assertGreaterEqual(result["test"][0]["test_avg_em"] , 0.2 )
@require_torch_gpu
@require_ray
def _UpperCamelCase( self : List[str] ):
a__ : int = self._run_finetune(gpus=1 , distributed_retriever="ray" )
self.assertGreaterEqual(result["test"][0]["test_avg_em"] , 0.2 )
@require_torch_multi_gpu
@require_ray
def _UpperCamelCase( self : List[Any] ):
a__ : str = self._run_finetune(gpus=1 , distributed_retriever="ray" )
self.assertGreaterEqual(result["test"][0]["test_avg_em"] , 0.2 )
| 37 |
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : List[str] ):
a__ : str = name
a__ : Optional[int] = value
a__ : Dict = weight
def __repr__( self : Union[str, Any] ):
return f'''{self.__class__.__name__}({self.name}, {self.value}, {self.weight})'''
def _UpperCamelCase( self : Dict ):
return self.value
def _UpperCamelCase( self : Optional[Any] ):
return self.name
def _UpperCamelCase( self : Optional[Any] ):
return self.weight
def _UpperCamelCase( self : Optional[int] ):
return self.value / self.weight
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = []
for i in range(len(__a ) ):
menu.append(Things(name[i] , value[i] , weight[i] ) )
return menu
def UpperCamelCase_ ( __a , __a , __a ) -> Union[str, Any]:
a__ : List[str] = sorted(__a , key=__a , reverse=__a )
a__ : List[Any] = []
a__, a__ : Union[str, Any] = 0.0, 0.0
for i in range(len(__a ) ):
if (total_cost + items_copy[i].get_weight()) <= max_cost:
result.append(items_copy[i] )
total_cost += items_copy[i].get_weight()
total_value += items_copy[i].get_value()
return (result, total_value)
def UpperCamelCase_ ( ) -> Union[str, Any]:
pass
if __name__ == "__main__":
import doctest
doctest.testmod()
| 37 | 1 |
def UpperCamelCase_ ( __a , __a ) -> int:
return x if y == 0 else greatest_common_divisor(__a , x % y )
def UpperCamelCase_ ( __a , __a ) -> int:
return (x * y) // greatest_common_divisor(__a , __a )
def UpperCamelCase_ ( __a = 20 ) -> int:
a__ : str = 1
for i in range(1 , n + 1 ):
a__ : List[Any] = lcm(__a , __a )
return g
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 |
import multiprocessing
from typing import TYPE_CHECKING, Optional, Union
from .. import Dataset, Features, config
from ..formatting import query_table
from ..packaged_modules.sql.sql import Sql
from ..utils import logging
from .abc import AbstractDatasetInputStream
if TYPE_CHECKING:
import sqlitea
import sqlalchemy
class A__ ( A__ ):
"""simple docstring"""
def __init__( self : Dict , lowerCamelCase__ : Union[str, "sqlalchemy.sql.Selectable"] , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[Features] = None , lowerCamelCase__ : str = None , lowerCamelCase__ : bool = False , **lowerCamelCase__ : Optional[int] , ):
super().__init__(features=lowerCamelCase__ , cache_dir=lowerCamelCase__ , keep_in_memory=lowerCamelCase__ , **lowerCamelCase__ )
a__ : str = Sql(
cache_dir=lowerCamelCase__ , features=lowerCamelCase__ , sql=lowerCamelCase__ , con=lowerCamelCase__ , **lowerCamelCase__ , )
def _UpperCamelCase( self : Tuple ):
a__ : Optional[Any] = None
a__ : Dict = None
a__ : Union[str, Any] = None
a__ : Union[str, Any] = None
self.builder.download_and_prepare(
download_config=lowerCamelCase__ , download_mode=lowerCamelCase__ , verification_mode=lowerCamelCase__ , base_path=lowerCamelCase__ , )
# Build dataset for splits
a__ : List[str] = self.builder.as_dataset(
split="train" , verification_mode=lowerCamelCase__ , in_memory=self.keep_in_memory )
return dataset
class A__ :
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : Dataset , lowerCamelCase__ : str , lowerCamelCase__ : Union[str, "sqlalchemy.engine.Connection", "sqlalchemy.engine.Engine", "sqlite3.Connection"] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[int] = None , **lowerCamelCase__ : Optional[Any] , ):
if num_proc is not None and num_proc <= 0:
raise ValueError(f'''num_proc {num_proc} must be an integer > 0.''' )
a__ : Any = dataset
a__ : str = name
a__ : Tuple = con
a__ : List[Any] = batch_size if batch_size else config.DEFAULT_MAX_BATCH_SIZE
a__ : Any = num_proc
a__ : Tuple = to_sql_kwargs
def _UpperCamelCase( self : List[Any] ):
a__ : Any = self.to_sql_kwargs.pop("sql" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("con" , lowerCamelCase__ )
a__ : int = self.to_sql_kwargs.pop("index" , lowerCamelCase__ )
a__ : int = self._write(index=lowerCamelCase__ , **self.to_sql_kwargs )
return written
def _UpperCamelCase( self : Any , lowerCamelCase__ : List[str] ):
a__, a__, a__ : Union[str, Any] = args
a__ : Any = {**to_sql_kwargs, "if_exists": "append"} if offset > 0 else to_sql_kwargs
a__ : Tuple = query_table(
table=self.dataset.data , key=slice(lowerCamelCase__ , offset + self.batch_size ) , indices=self.dataset._indices , )
a__ : str = batch.to_pandas()
a__ : List[Any] = df.to_sql(self.name , self.con , index=lowerCamelCase__ , **lowerCamelCase__ )
return num_rows or len(lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Optional[int] , **lowerCamelCase__ : Optional[Any] ):
a__ : str = 0
if self.num_proc is None or self.num_proc == 1:
for offset in logging.tqdm(
range(0 , len(self.dataset ) , self.batch_size ) , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += self._batch_sql((offset, index, to_sql_kwargs) )
else:
a__, a__ : List[str] = len(self.dataset ), self.batch_size
with multiprocessing.Pool(self.num_proc ) as pool:
for num_rows in logging.tqdm(
pool.imap(
self._batch_sql , [(offset, index, to_sql_kwargs) for offset in range(0 , lowerCamelCase__ , lowerCamelCase__ )] , ) , total=(num_rows // batch_size) + 1 if num_rows % batch_size else num_rows // batch_size , unit="ba" , disable=not logging.is_progress_bar_enabled() , desc="Creating SQL from Arrow format" , ):
written += num_rows
return written
| 37 | 1 |
import argparse
import json
import os
import pickle
import shutil
import numpy as np
import torch
from distiller import Distiller
from lm_seqs_dataset import LmSeqsDataset
from transformers import (
BertConfig,
BertForMaskedLM,
BertTokenizer,
DistilBertConfig,
DistilBertForMaskedLM,
DistilBertTokenizer,
GPTaConfig,
GPTaLMHeadModel,
GPTaTokenizer,
RobertaConfig,
RobertaForMaskedLM,
RobertaTokenizer,
)
from utils import git_log, init_gpu_params, logger, set_seed
UpperCamelCase : int = {
"""distilbert""": (DistilBertConfig, DistilBertForMaskedLM, DistilBertTokenizer),
"""roberta""": (RobertaConfig, RobertaForMaskedLM, RobertaTokenizer),
"""bert""": (BertConfig, BertForMaskedLM, BertTokenizer),
"""gpt2""": (GPTaConfig, GPTaLMHeadModel, GPTaTokenizer),
}
def UpperCamelCase_ ( __a ) -> List[Any]:
assert (args.mlm and args.alpha_mlm > 0.0) or (not args.mlm and args.alpha_mlm == 0.0)
assert (args.alpha_mlm > 0.0 and args.alpha_clm == 0.0) or (args.alpha_mlm == 0.0 and args.alpha_clm > 0.0)
if args.mlm:
assert os.path.isfile(args.token_counts )
assert (args.student_type in ["roberta", "distilbert"]) and (args.teacher_type in ["roberta", "bert"])
else:
assert (args.student_type in ["gpt2"]) and (args.teacher_type in ["gpt2"])
assert args.teacher_type == args.student_type or (
args.student_type == "distilbert" and args.teacher_type == "bert"
)
assert os.path.isfile(args.student_config )
if args.student_pretrained_weights is not None:
assert os.path.isfile(args.student_pretrained_weights )
if args.freeze_token_type_embds:
assert args.student_type in ["roberta"]
assert args.alpha_ce >= 0.0
assert args.alpha_mlm >= 0.0
assert args.alpha_clm >= 0.0
assert args.alpha_mse >= 0.0
assert args.alpha_cos >= 0.0
assert args.alpha_ce + args.alpha_mlm + args.alpha_clm + args.alpha_mse + args.alpha_cos > 0.0
def UpperCamelCase_ ( __a , __a ) -> int:
if args.student_type == "roberta":
a__ : str = False
elif args.student_type == "gpt2":
a__ : List[str] = False
def UpperCamelCase_ ( __a , __a ) -> Optional[int]:
if args.student_type == "roberta":
a__ : Union[str, Any] = False
def UpperCamelCase_ ( ) -> Any:
a__ : List[Any] = argparse.ArgumentParser(description="Training" )
parser.add_argument("--force" , action="store_true" , help="Overwrite dump_path if it already exists." )
parser.add_argument(
"--dump_path" , type=__a , required=__a , help="The output directory (log, checkpoints, parameters, etc.)" )
parser.add_argument(
"--data_file" , type=__a , required=__a , help="The binarized file (tokenized + tokens_to_ids) and grouped by sequence." , )
parser.add_argument(
"--student_type" , type=__a , choices=["distilbert", "roberta", "gpt2"] , required=__a , help="The student type (DistilBERT, RoBERTa)." , )
parser.add_argument("--student_config" , type=__a , required=__a , help="Path to the student configuration." )
parser.add_argument(
"--student_pretrained_weights" , default=__a , type=__a , help="Load student initialization checkpoint." )
parser.add_argument(
"--teacher_type" , choices=["bert", "roberta", "gpt2"] , required=__a , help="Teacher type (BERT, RoBERTa)." )
parser.add_argument("--teacher_name" , type=__a , required=__a , help="The teacher model." )
parser.add_argument("--temperature" , default=2.0 , type=__a , help="Temperature for the softmax temperature." )
parser.add_argument(
"--alpha_ce" , default=0.5 , type=__a , help="Linear weight for the distillation loss. Must be >=0." )
parser.add_argument(
"--alpha_mlm" , default=0.0 , type=__a , help="Linear weight for the MLM loss. Must be >=0. Should be used in conjunction with `mlm` flag." , )
parser.add_argument("--alpha_clm" , default=0.5 , type=__a , help="Linear weight for the CLM loss. Must be >=0." )
parser.add_argument("--alpha_mse" , default=0.0 , type=__a , help="Linear weight of the MSE loss. Must be >=0." )
parser.add_argument(
"--alpha_cos" , default=0.0 , type=__a , help="Linear weight of the cosine embedding loss. Must be >=0." )
parser.add_argument(
"--mlm" , action="store_true" , help="The LM step: MLM or CLM. If `mlm` is True, the MLM is used over CLM." )
parser.add_argument(
"--mlm_mask_prop" , default=0.15 , type=__a , help="Proportion of tokens for which we need to make a prediction." , )
parser.add_argument("--word_mask" , default=0.8 , type=__a , help="Proportion of tokens to mask out." )
parser.add_argument("--word_keep" , default=0.1 , type=__a , help="Proportion of tokens to keep." )
parser.add_argument("--word_rand" , default=0.1 , type=__a , help="Proportion of tokens to randomly replace." )
parser.add_argument(
"--mlm_smoothing" , default=0.7 , type=__a , help="Smoothing parameter to emphasize more rare tokens (see XLM, similar to word2vec)." , )
parser.add_argument("--token_counts" , type=__a , help="The token counts in the data_file for MLM." )
parser.add_argument(
"--restrict_ce_to_mask" , action="store_true" , help="If true, compute the distillation loss only the [MLM] prediction distribution." , )
parser.add_argument(
"--freeze_pos_embs" , action="store_true" , help="Freeze positional embeddings during distillation. For student_type in ['roberta', 'gpt2'] only." , )
parser.add_argument(
"--freeze_token_type_embds" , action="store_true" , help="Freeze token type embeddings during distillation if existent. For student_type in ['roberta'] only." , )
parser.add_argument("--n_epoch" , type=__a , default=3 , help="Number of pass on the whole dataset." )
parser.add_argument("--batch_size" , type=__a , default=5 , help="Batch size (for each process)." )
parser.add_argument(
"--group_by_size" , action="store_false" , help="If true, group sequences that have similar length into the same batch. Default is true." , )
parser.add_argument(
"--gradient_accumulation_steps" , type=__a , default=50 , help="Gradient accumulation for larger training batches." , )
parser.add_argument("--warmup_prop" , default=0.05 , type=__a , help="Linear warmup proportion." )
parser.add_argument("--weight_decay" , default=0.0 , type=__a , help="Weight decay if we apply some." )
parser.add_argument("--learning_rate" , default=5e-4 , type=__a , help="The initial learning rate for Adam." )
parser.add_argument("--adam_epsilon" , default=1e-6 , type=__a , help="Epsilon for Adam optimizer." )
parser.add_argument("--max_grad_norm" , default=5.0 , type=__a , help="Max gradient norm." )
parser.add_argument("--initializer_range" , default=0.02 , type=__a , help="Random initialization range." )
parser.add_argument(
"--fp16" , action="store_true" , help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit" , )
parser.add_argument(
"--fp16_opt_level" , type=__a , default="O1" , help=(
"For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
"See details at https://nvidia.github.io/apex/amp.html"
) , )
parser.add_argument("--n_gpu" , type=__a , default=1 , help="Number of GPUs in the node." )
parser.add_argument("--local_rank" , type=__a , default=-1 , help="Distributed training - Local rank" )
parser.add_argument("--seed" , type=__a , default=56 , help="Random seed" )
parser.add_argument("--log_interval" , type=__a , default=500 , help="Tensorboard logging interval." )
parser.add_argument("--checkpoint_interval" , type=__a , default=4_000 , help="Checkpoint interval." )
a__ : List[str] = parser.parse_args()
sanity_checks(__a )
# ARGS #
init_gpu_params(__a )
set_seed(__a )
if args.is_master:
if os.path.exists(args.dump_path ):
if not args.force:
raise ValueError(
f'''Serialization dir {args.dump_path} already exists, but you have not precised wheter to overwrite'''
" itUse `--force` if you want to overwrite it" )
else:
shutil.rmtree(args.dump_path )
if not os.path.exists(args.dump_path ):
os.makedirs(args.dump_path )
logger.info(f'''Experiment will be dumped and logged in {args.dump_path}''' )
# SAVE PARAMS #
logger.info(f'''Param: {args}''' )
with open(os.path.join(args.dump_path , "parameters.json" ) , "w" ) as f:
json.dump(vars(__a ) , __a , indent=4 )
git_log(args.dump_path )
a__, a__, a__ : Dict = MODEL_CLASSES[args.student_type]
a__, a__, a__ : List[Any] = MODEL_CLASSES[args.teacher_type]
# TOKENIZER #
a__ : Optional[int] = teacher_tokenizer_class.from_pretrained(args.teacher_name )
a__ : Union[str, Any] = {}
for tok_name, tok_symbol in tokenizer.special_tokens_map.items():
a__ : Optional[Any] = tokenizer.all_special_tokens.index(__a )
a__ : Union[str, Any] = tokenizer.all_special_ids[idx]
logger.info(f'''Special tokens {special_tok_ids}''' )
a__ : List[str] = special_tok_ids
a__ : Optional[Any] = tokenizer.max_model_input_sizes[args.teacher_name]
# DATA LOADER #
logger.info(f'''Loading data from {args.data_file}''' )
with open(args.data_file , "rb" ) as fp:
a__ : Tuple = pickle.load(__a )
if args.mlm:
logger.info(f'''Loading token counts from {args.token_counts} (already pre-computed)''' )
with open(args.token_counts , "rb" ) as fp:
a__ : Optional[Any] = pickle.load(__a )
a__ : Dict = np.maximum(__a , 1 ) ** -args.mlm_smoothing
for idx in special_tok_ids.values():
a__ : List[Any] = 0.0 # do not predict special tokens
a__ : List[str] = torch.from_numpy(__a )
else:
a__ : Any = None
a__ : Any = LmSeqsDataset(params=__a , data=__a )
logger.info("Data loader created." )
# STUDENT #
logger.info(f'''Loading student config from {args.student_config}''' )
a__ : List[str] = student_config_class.from_pretrained(args.student_config )
a__ : Optional[Any] = True
if args.student_pretrained_weights is not None:
logger.info(f'''Loading pretrained weights from {args.student_pretrained_weights}''' )
a__ : List[Any] = student_model_class.from_pretrained(args.student_pretrained_weights , config=__a )
else:
a__ : List[Any] = student_model_class(__a )
if args.n_gpu > 0:
student.to(f'''cuda:{args.local_rank}''' )
logger.info("Student loaded." )
# TEACHER #
a__ : int = teacher_model_class.from_pretrained(args.teacher_name , output_hidden_states=__a )
if args.n_gpu > 0:
teacher.to(f'''cuda:{args.local_rank}''' )
logger.info(f'''Teacher loaded from {args.teacher_name}.''' )
# FREEZING #
if args.freeze_pos_embs:
freeze_pos_embeddings(__a , __a )
if args.freeze_token_type_embds:
freeze_token_type_embeddings(__a , __a )
# SANITY CHECKS #
assert student.config.vocab_size == teacher.config.vocab_size
assert student.config.hidden_size == teacher.config.hidden_size
assert student.config.max_position_embeddings == teacher.config.max_position_embeddings
if args.mlm:
assert token_probs.size(0 ) == stu_architecture_config.vocab_size
# DISTILLER #
torch.cuda.empty_cache()
a__ : List[Any] = Distiller(
params=__a , dataset=__a , token_probs=__a , student=__a , teacher=__a )
distiller.train()
logger.info("Let's go get some drinks." )
if __name__ == "__main__":
main()
| 37 |
import math
from datetime import datetime, timedelta
def UpperCamelCase_ ( __a ) -> datetime:
a__ : Union[str, Any] = year % 19
a__ : List[str] = year % 4
a__ : str = year % 7
a__ : Any = math.floor(year / 100 )
a__ : List[str] = math.floor((13 + 8 * leap_day_inhibits) / 25 )
a__ : Optional[int] = leap_day_inhibits / 4
a__ : Union[str, Any] = (
15 - lunar_orbit_correction + leap_day_inhibits - leap_day_reinstall_number
) % 30
a__ : Dict = (4 + leap_day_inhibits - leap_day_reinstall_number) % 7
# days to be added to March 21
a__ : Any = (19 * metonic_cycle + secular_moon_shift) % 30
# PHM -> Paschal Full Moon
a__ : List[Any] = (
2 * julian_leap_year
+ 4 * non_leap_year
+ 6 * days_to_add
+ century_starting_point
) % 7
if days_to_add == 29 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 19 )
elif days_to_add == 28 and days_from_phm_to_sunday == 6:
return datetime(__a , 4 , 18 )
else:
return datetime(__a , 3 , 22 ) + timedelta(
days=int(days_to_add + days_from_phm_to_sunday ) )
if __name__ == "__main__":
for year in (1994, 2000, 2010, 2021, 2023):
UpperCamelCase : Tuple = """will be""" if year > datetime.now().year else """was"""
print(f"""Easter in {year} {tense} {gauss_easter(year)}""")
| 37 | 1 |
import argparse
from transformers import TaConfig, TaForConditionalGeneration, load_tf_weights_in_ta
from transformers.utils import logging
logging.set_verbosity_info()
def UpperCamelCase_ ( __a , __a , __a ) -> Union[str, Any]:
# Initialise PyTorch model
a__ : List[Any] = TaConfig.from_json_file(__a )
print(f'''Building PyTorch model from configuration: {config}''' )
a__ : Optional[int] = TaForConditionalGeneration(__a )
# Load weights from tf checkpoint
load_tf_weights_in_ta(__a , __a , __a )
# Save pytorch-model
print(f'''Save PyTorch model to {pytorch_dump_path}''' )
model.save_pretrained(__a )
if __name__ == "__main__":
UpperCamelCase : Tuple = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"""--tf_checkpoint_path""", default=None, type=str, required=True, help="""Path to the TensorFlow checkpoint path."""
)
parser.add_argument(
"""--config_file""",
default=None,
type=str,
required=True,
help=(
"""The config json file corresponding to the pre-trained T5 model. \nThis specifies the model architecture."""
),
)
parser.add_argument(
"""--pytorch_dump_path""", default=None, type=str, required=True, help="""Path to the output PyTorch model."""
)
UpperCamelCase : Tuple = parser.parse_args()
convert_tf_checkpoint_to_pytorch(args.tf_checkpoint_path, args.config_file, args.pytorch_dump_path)
| 37 |
import gc
import importlib.metadata
import tempfile
import unittest
from packaging import version
from transformers import (
AutoModel,
AutoModelForCausalLM,
AutoModelForSeqaSeqLM,
AutoModelForSequenceClassification,
AutoTokenizer,
BitsAndBytesConfig,
pipeline,
)
from transformers.testing_utils import (
is_torch_available,
require_accelerate,
require_bitsandbytes,
require_torch,
require_torch_gpu,
require_torch_multi_gpu,
slow,
)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
if model.config.model_type == "gpt2":
return model.transformer.h[0].mlp.c_fc
return model.transformer.h[0].mlp.dense_ah_to_h
if is_torch_available():
import torch
import torch.nn as nn
class A__ ( nn.Module ):
"""simple docstring"""
def __init__( self : List[str] , lowerCamelCase__ : nn.Module , lowerCamelCase__ : int ):
super().__init__()
a__ : int = module
a__ : Any = nn.Sequential(
nn.Linear(module.in_features , lowerCamelCase__ , bias=lowerCamelCase__ ) , nn.Linear(lowerCamelCase__ , module.out_features , bias=lowerCamelCase__ ) , )
a__ : Tuple = (2.0 / (5 * min(module.in_features , module.out_features ))) ** 0.5
nn.init.normal_(self.adapter[0].weight , std=lowerCamelCase__ )
nn.init.zeros_(self.adapter[1].weight )
self.adapter.to(module.weight.device )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Optional[int] , *lowerCamelCase__ : int , **lowerCamelCase__ : Dict ):
return self.module(lowerCamelCase__ , *lowerCamelCase__ , **lowerCamelCase__ ) + self.adapter(lowerCamelCase__ )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
_lowercase = 'bigscience/bloom-1b7'
# Constant values
_lowercase = 2.1_09_65_95_52_69_25_74
_lowercase = 'Hello my name is'
_lowercase = set()
EXPECTED_OUTPUTS.add('Hello my name is John and I am a professional photographer. I' )
EXPECTED_OUTPUTS.add('Hello my name is John.\nI am a friend of your father.\n' )
EXPECTED_OUTPUTS.add('Hello my name is John Doe, I am a student at the University' )
_lowercase = 1_0
def _UpperCamelCase( self : Dict ):
# Models and tokenizer
a__ : List[str] = AutoTokenizer.from_pretrained(self.model_name )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Union[str, Any] ):
super().setUp()
# Models and tokenizer
a__ : List[Any] = AutoModelForCausalLM.from_pretrained(
self.model_name , torch_dtype=torch.floataa , device_map="auto" )
a__ : Union[str, Any] = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.model_fpaa
del self.model_abit
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : List[Any] ):
a__ : str = self.model_abit.config
self.assertTrue(hasattr(lowerCamelCase__ , "quantization_config" ) )
a__ : Optional[Any] = config.to_dict()
a__ : int = config.to_diff_dict()
a__ : List[str] = config.to_json_string()
def _UpperCamelCase( self : int ):
from bitsandbytes.nn import Paramsabit
a__ : List[Any] = self.model_fpaa.get_memory_footprint()
a__ : str = self.model_abit.get_memory_footprint()
self.assertAlmostEqual(mem_fpaa / mem_abit , self.EXPECTED_RELATIVE_DIFFERENCE )
a__ : Optional[Any] = get_some_linear_layer(self.model_abit )
self.assertTrue(linear.weight.__class__ == Paramsabit )
def _UpperCamelCase( self : Tuple ):
from transformers import TaPreTrainedModel
self.model_fpaa.get_memory_footprint()
self.model_abit.get_memory_footprint()
for name, module in self.model_abit.named_modules():
if isinstance(lowerCamelCase__ , torch.nn.Linear ):
if name not in ["lm_head"] + TaPreTrainedModel._keep_in_fpaa_modules:
# 4-bit parameters are packed in uint8 variables
self.assertTrue(module.weight.dtype == torch.uinta )
def _UpperCamelCase( self : str ):
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Tuple = self.model_abit.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[Any] = BitsAndBytesConfig()
a__ : Optional[int] = True
a__ : int = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , device_map="auto" )
a__ : str = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : int = model_abit_from_config.generate(
input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_sequences[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
def _UpperCamelCase( self : Dict ):
with self.assertRaises(lowerCamelCase__ ), tempfile.TemporaryDirectory() as tmpdirname:
self.model_abit.save_pretrained(lowerCamelCase__ )
def _UpperCamelCase( self : Union[str, Any] ):
a__ : int = BitsAndBytesConfig()
with self.assertRaises(lowerCamelCase__ ):
a__ : Dict = AutoModelForCausalLM.from_pretrained(
self.model_name , quantization_config=lowerCamelCase__ , load_in_abit=lowerCamelCase__ , device_map="auto" , bnb_abit_quant_type="nf4" , )
def _UpperCamelCase( self : int ):
with self.assertRaises(lowerCamelCase__ ):
# Tries with `str`
self.model_abit.to("cpu" )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `dtype``
self.model_abit.to(torch.floataa )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.to(torch.device("cuda:0" ) )
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.float()
with self.assertRaises(lowerCamelCase__ ):
# Tries with a `device`
self.model_abit.half()
# Test if we did not break anything
a__ : int = self.tokenizer(self.input_text , return_tensors="pt" )
a__ : Any = self.model_fpaa.to(torch.floataa )
a__ : List[Any] = self.model_fpaa.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.to("cpu" )
# Check this does not throw an error
a__ : Tuple = self.model_fpaa.half()
# Check this does not throw an error
a__ : Dict = self.model_fpaa.float()
def _UpperCamelCase( self : Dict ):
a__ : List[str] = AutoModelForSeqaSeqLM.from_pretrained("t5-small" , load_in_abit=lowerCamelCase__ , device_map="auto" )
self.assertTrue(model.decoder.block[0].layer[2].DenseReluDense.wo.weight.dtype == torch.floataa )
@require_bitsandbytes
@require_accelerate
@require_torch
@require_torch_gpu
@slow
class A__ ( unittest.TestCase ):
"""simple docstring"""
@classmethod
def _UpperCamelCase( cls : str ):
a__ : Dict = "t5-small"
a__ : List[Any] = "google/flan-t5-small" # flan-t5 uses dense-act instead of dense-relu-dense
a__ : int = AutoTokenizer.from_pretrained(cls.model_name )
a__ : str = "Translate in German: Hello, my dog is cute"
def _UpperCamelCase( self : Optional[int] ):
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Optional[int] ):
from transformers import TaForConditionalGeneration
a__ : List[Any] = TaForConditionalGeneration._keep_in_fpaa_modules
a__ : Optional[Any] = None
# test with `t5-small`
a__ : Any = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : Dict = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Dict = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Any = model.generate(**lowerCamelCase__ )
a__ : Union[str, Any] = modules
def _UpperCamelCase( self : List[Any] ):
import bitsandbytes as bnb
from transformers import TaForConditionalGeneration
# test with `t5-small`
a__ : List[str] = TaForConditionalGeneration.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# there was a bug with decoders - this test checks that it is fixed
self.assertTrue(isinstance(model.decoder.block[0].layer[0].SelfAttention.q , bnb.nn.Linearabit ) )
a__ : Union[str, Any] = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : int = model.generate(**lowerCamelCase__ )
# test with `flan-t5-small`
a__ : int = TaForConditionalGeneration.from_pretrained(
self.dense_act_model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
a__ : Any = self.tokenizer(self.input_text , return_tensors="pt" ).to(0 )
a__ : Optional[int] = model.generate(**lowerCamelCase__ )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : List[str] ):
super().setUp()
# model_name
a__ : Union[str, Any] = "bigscience/bloom-560m"
a__ : Union[str, Any] = "t5-small"
# Different types of model
a__ : int = AutoModel.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Sequence classification model
a__ : Dict = AutoModelForSequenceClassification.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# CausalLM model
a__ : str = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
# Seq2seq model
a__ : Dict = AutoModelForSeqaSeqLM.from_pretrained(
self.seq_to_seq_name , load_in_abit=lowerCamelCase__ , device_map="auto" )
def _UpperCamelCase( self : List[Any] ):
del self.base_model
del self.sequence_model
del self.model_abit
del self.seq_to_seq_model
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Union[str, Any] ):
from bitsandbytes.nn import Paramsabit
self.assertTrue(self.base_model.h[-1].mlp.dense_ah_to_h.weight.__class__ == Paramsabit )
# Other heads should be nn.Parameter
self.assertTrue(self.model_abit.lm_head.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.sequence_model.score.weight.__class__ == torch.nn.Parameter )
self.assertTrue(self.seq_to_seq_model.lm_head.weight.__class__ == torch.nn.Parameter )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
super().setUp()
def _UpperCamelCase( self : int ):
del self.pipe
gc.collect()
torch.cuda.empty_cache()
def _UpperCamelCase( self : Tuple ):
a__ : int = pipeline(
"text-generation" , model=self.model_name , model_kwargs={"device_map": "auto", "load_in_4bit": True, "torch_dtype": torch.floataa} , max_new_tokens=self.MAX_NEW_TOKENS , )
# Real second forward pass
a__ : Tuple = self.pipe(self.input_text )
self.assertIn(pipeline_output[0]["generated_text"] , self.EXPECTED_OUTPUTS )
@require_torch_multi_gpu
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Tuple ):
super().setUp()
def _UpperCamelCase( self : List[Any] ):
a__ : str = AutoModelForCausalLM.from_pretrained(
self.model_name , load_in_abit=lowerCamelCase__ , device_map="balanced" )
# Check correct device map
self.assertEqual(set(model_parallel.hf_device_map.values() ) , {0, 1} )
# Check that inference pass works on the model
a__ : List[Any] = self.tokenizer(self.input_text , return_tensors="pt" )
# Second real batch
a__ : List[Any] = model_parallel.generate(input_ids=encoded_input["input_ids"].to(0 ) , max_new_tokens=10 )
self.assertIn(self.tokenizer.decode(output_parallel[0] , skip_special_tokens=lowerCamelCase__ ) , self.EXPECTED_OUTPUTS )
class A__ ( A__ ):
"""simple docstring"""
def _UpperCamelCase( self : Dict ):
a__ : Any = "facebook/opt-350m"
super().setUp()
def _UpperCamelCase( self : int ):
if version.parse(importlib.metadata.version("bitsandbytes" ) ) < version.parse("0.37.0" ):
return
# Step 1: freeze all parameters
a__ : Tuple = AutoModelForCausalLM.from_pretrained(self.model_name , load_in_abit=lowerCamelCase__ )
self.assertEqual(set(model.hf_device_map.values() ) , {torch.cuda.current_device()} )
for param in model.parameters():
a__ : Any = False # freeze the model - train adapters later
if param.ndim == 1:
# cast the small parameters (e.g. layernorm) to fp32 for stability
a__ : Tuple = param.data.to(torch.floataa )
# Step 2: add adapters
for _, module in model.named_modules():
if "OPTAttention" in repr(type(lowerCamelCase__ ) ):
a__ : Dict = LoRALayer(module.q_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.k_proj , rank=16 )
a__ : List[Any] = LoRALayer(module.v_proj , rank=16 )
# Step 3: dummy batch
a__ : Dict = self.tokenizer("Test batch " , return_tensors="pt" ).to(0 )
# Step 4: Check if the gradient is not None
with torch.cuda.amp.autocast():
a__ : Optional[Any] = model.forward(**lowerCamelCase__ )
out.logits.norm().backward()
for module in model.modules():
if isinstance(lowerCamelCase__ , lowerCamelCase__ ):
self.assertTrue(module.adapter[1].weight.grad is not None )
self.assertTrue(module.adapter[1].weight.grad.norm().item() > 0 )
elif isinstance(lowerCamelCase__ , nn.Embedding ):
self.assertTrue(module.weight.grad is None )
class A__ ( A__ ):
"""simple docstring"""
_lowercase = 'gpt2-xl'
_lowercase = 3.31_91_85_48_54_15_21_87
| 37 | 1 |
import json
import os
from functools import lru_cache
from typing import Dict, List, Optional, Tuple, Union
import regex as re
from ...tokenization_utils import AddedToken, PreTrainedTokenizer
from ...tokenization_utils_base import BatchEncoding, EncodedInput
from ...utils import PaddingStrategy, logging
UpperCamelCase : List[str] = logging.get_logger(__name__)
UpperCamelCase : Optional[int] = {"""vocab_file""": """vocab.json""", """merges_file""": """merges.txt"""}
# See all LED models at https://huggingface.co/models?filter=LED
UpperCamelCase : List[Any] = {
"""vocab_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/vocab.json""",
},
"""merges_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/merges.txt""",
},
"""tokenizer_file""": {
"""allenai/led-base-16384""": """https://huggingface.co/allenai/led-base-16384/resolve/main/tokenizer.json""",
},
}
UpperCamelCase : Optional[Any] = {
"""allenai/led-base-16384""": 1_6384,
}
@lru_cache()
# Copied from transformers.models.bart.tokenization_bart.bytes_to_unicode
def UpperCamelCase_ ( ) -> Union[str, Any]:
a__ : Optional[int] = (
list(range(ord("!" ) , ord("~" ) + 1 ) ) + list(range(ord("¡" ) , ord("¬" ) + 1 ) ) + list(range(ord("®" ) , ord("ÿ" ) + 1 ) )
)
a__ : Dict = bs[:]
a__ : Tuple = 0
for b in range(2**8 ):
if b not in bs:
bs.append(__a )
cs.append(2**8 + n )
n += 1
a__ : Any = [chr(__a ) for n in cs]
return dict(zip(__a , __a ) )
def UpperCamelCase_ ( __a ) -> Optional[int]:
a__ : Tuple = set()
a__ : List[str] = word[0]
for char in word[1:]:
pairs.add((prev_char, char) )
a__ : Dict = char
return pairs
class A__ ( A__ ):
"""simple docstring"""
_lowercase = VOCAB_FILES_NAMES
_lowercase = PRETRAINED_VOCAB_FILES_MAP
_lowercase = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
_lowercase = ['input_ids', 'attention_mask']
def __init__( self : Tuple , lowerCamelCase__ : List[str] , lowerCamelCase__ : Any , lowerCamelCase__ : Dict="replace" , lowerCamelCase__ : int="<s>" , lowerCamelCase__ : Tuple="</s>" , lowerCamelCase__ : Optional[int]="</s>" , lowerCamelCase__ : Dict="<s>" , lowerCamelCase__ : Union[str, Any]="<unk>" , lowerCamelCase__ : Optional[int]="<pad>" , lowerCamelCase__ : Dict="<mask>" , lowerCamelCase__ : Optional[int]=False , **lowerCamelCase__ : int , ):
a__ : Optional[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else bos_token
a__ : List[str] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else eos_token
a__ : Dict = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else sep_token
a__ : Optional[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else cls_token
a__ : Optional[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else unk_token
a__ : Tuple = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else pad_token
# Mask token behave like a normal word, i.e. include the space before it
a__ : List[Any] = AddedToken(lowerCamelCase__ , lstrip=lowerCamelCase__ , rstrip=lowerCamelCase__ ) if isinstance(lowerCamelCase__ , lowerCamelCase__ ) else mask_token
super().__init__(
errors=lowerCamelCase__ , bos_token=lowerCamelCase__ , eos_token=lowerCamelCase__ , unk_token=lowerCamelCase__ , sep_token=lowerCamelCase__ , cls_token=lowerCamelCase__ , pad_token=lowerCamelCase__ , mask_token=lowerCamelCase__ , add_prefix_space=lowerCamelCase__ , **lowerCamelCase__ , )
with open(lowerCamelCase__ , encoding="utf-8" ) as vocab_handle:
a__ : List[str] = json.load(lowerCamelCase__ )
a__ : List[str] = {v: k for k, v in self.encoder.items()}
a__ : int = errors # how to handle errors in decoding
a__ : List[str] = bytes_to_unicode()
a__ : int = {v: k for k, v in self.byte_encoder.items()}
with open(lowerCamelCase__ , encoding="utf-8" ) as merges_handle:
a__ : List[str] = merges_handle.read().split("\n" )[1:-1]
a__ : Optional[int] = [tuple(merge.split() ) for merge in bpe_merges]
a__ : Tuple = dict(zip(lowerCamelCase__ , range(len(lowerCamelCase__ ) ) ) )
a__ : Union[str, Any] = {}
a__ : Optional[int] = add_prefix_space
# Should have added re.IGNORECASE so BPE merges can happen for capitalized versions of contractions
a__ : Dict = re.compile(r"'s|'t|'re|'ve|'m|'ll|'d| ?\p{L}+| ?\p{N}+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+" )
@property
# Copied from transformers.models.bart.tokenization_bart.BartTokenizer.vocab_size
def _UpperCamelCase( self : Dict ):
return len(self.encoder )
def _UpperCamelCase( self : Tuple ):
return dict(self.encoder , **self.added_tokens_encoder )
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : Union[str, Any] ):
if token in self.cache:
return self.cache[token]
a__ : str = tuple(lowerCamelCase__ )
a__ : Dict = get_pairs(lowerCamelCase__ )
if not pairs:
return token
while True:
a__ : List[str] = min(lowerCamelCase__ , key=lambda lowerCamelCase__ : self.bpe_ranks.get(lowerCamelCase__ , float("inf" ) ) )
if bigram not in self.bpe_ranks:
break
a__, a__ : Union[str, Any] = bigram
a__ : Any = []
a__ : Union[str, Any] = 0
while i < len(lowerCamelCase__ ):
try:
a__ : Dict = word.index(lowerCamelCase__ , lowerCamelCase__ )
except ValueError:
new_word.extend(word[i:] )
break
else:
new_word.extend(word[i:j] )
a__ : List[str] = j
if word[i] == first and i < len(lowerCamelCase__ ) - 1 and word[i + 1] == second:
new_word.append(first + second )
i += 2
else:
new_word.append(word[i] )
i += 1
a__ : Tuple = tuple(lowerCamelCase__ )
a__ : Union[str, Any] = new_word
if len(lowerCamelCase__ ) == 1:
break
else:
a__ : str = get_pairs(lowerCamelCase__ )
a__ : Union[str, Any] = " ".join(lowerCamelCase__ )
a__ : Union[str, Any] = word
return word
def _UpperCamelCase( self : int , lowerCamelCase__ : Tuple ):
a__ : Optional[int] = []
for token in re.findall(self.pat , lowerCamelCase__ ):
a__ : Union[str, Any] = "".join(
self.byte_encoder[b] for b in token.encode("utf-8" ) ) # Maps all our bytes to unicode strings, avoiding control tokens of the BPE (spaces in our case)
bpe_tokens.extend(bpe_token for bpe_token in self.bpe(lowerCamelCase__ ).split(" " ) )
return bpe_tokens
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : Any ):
return self.encoder.get(lowerCamelCase__ , self.encoder.get(self.unk_token ) )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : str ):
return self.decoder.get(lowerCamelCase__ )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : Any ):
a__ : Optional[Any] = "".join(lowerCamelCase__ )
a__ : Optional[int] = bytearray([self.byte_decoder[c] for c in text] ).decode("utf-8" , errors=self.errors )
return text
def _UpperCamelCase( self : Union[str, Any] , lowerCamelCase__ : str , lowerCamelCase__ : Optional[str] = None ):
if not os.path.isdir(lowerCamelCase__ ):
logger.error(f'''Vocabulary path ({save_directory}) should be a directory''' )
return
a__ : Dict = os.path.join(
lowerCamelCase__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] )
a__ : List[Any] = os.path.join(
lowerCamelCase__ , (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["merges_file"] )
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as f:
f.write(json.dumps(self.encoder , indent=2 , sort_keys=lowerCamelCase__ , ensure_ascii=lowerCamelCase__ ) + "\n" )
a__ : Dict = 0
with open(lowerCamelCase__ , "w" , encoding="utf-8" ) as writer:
writer.write("#version: 0.2\n" )
for bpe_tokens, token_index in sorted(self.bpe_ranks.items() , key=lambda lowerCamelCase__ : kv[1] ):
if index != token_index:
logger.warning(
f'''Saving vocabulary to {merge_file}: BPE merge indices are not consecutive.'''
" Please check that the tokenizer is not corrupted!" )
a__ : List[Any] = token_index
writer.write(" ".join(lowerCamelCase__ ) + "\n" )
index += 1
return vocab_file, merge_file
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
if token_ids_a is None:
return [self.cls_token_id] + token_ids_a + [self.sep_token_id]
a__ : str = [self.cls_token_id]
a__ : List[str] = [self.sep_token_id]
return cls + token_ids_a + sep + sep + token_ids_a + sep
def _UpperCamelCase( self : Any , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None , lowerCamelCase__ : bool = False ):
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_a=lowerCamelCase__ , token_ids_a=lowerCamelCase__ , already_has_special_tokens=lowerCamelCase__ )
if token_ids_a is None:
return [1] + ([0] * len(lowerCamelCase__ )) + [1]
return [1] + ([0] * len(lowerCamelCase__ )) + [1, 1] + ([0] * len(lowerCamelCase__ )) + [1]
def _UpperCamelCase( self : List[str] , lowerCamelCase__ : List[int] , lowerCamelCase__ : Optional[List[int]] = None ):
a__ : int = [self.sep_token_id]
a__ : Union[str, Any] = [self.cls_token_id]
if token_ids_a is None:
return len(cls + token_ids_a + sep ) * [0]
return len(cls + token_ids_a + sep + sep + token_ids_a + sep ) * [0]
def _UpperCamelCase( self : Optional[int] , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Optional[int]=False , **lowerCamelCase__ : List[Any] ):
a__ : List[str] = kwargs.pop("add_prefix_space" , self.add_prefix_space )
if (is_split_into_words or add_prefix_space) and (len(lowerCamelCase__ ) > 0 and not text[0].isspace()):
a__ : List[Any] = " " + text
return (text, kwargs)
def _UpperCamelCase( self : Optional[Any] , lowerCamelCase__ : Union[Dict[str, EncodedInput], BatchEncoding] , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : PaddingStrategy = PaddingStrategy.DO_NOT_PAD , lowerCamelCase__ : Optional[int] = None , lowerCamelCase__ : Optional[bool] = None , ):
a__ : Dict = super()._pad(
encoded_inputs=lowerCamelCase__ , max_length=lowerCamelCase__ , padding_strategy=lowerCamelCase__ , pad_to_multiple_of=lowerCamelCase__ , return_attention_mask=lowerCamelCase__ , )
# Load from model defaults
if return_attention_mask is None:
a__ : int = "attention_mask" in self.model_input_names
if return_attention_mask and "global_attention_mask" in encoded_inputs:
a__ : Optional[int] = encoded_inputs[self.model_input_names[0]]
# `global_attention_mask` need to have the same length as other (sequential) inputs.
a__ : Optional[Any] = len(encoded_inputs["global_attention_mask"] ) != len(lowerCamelCase__ )
if needs_to_be_padded:
a__ : int = len(lowerCamelCase__ ) - len(encoded_inputs["global_attention_mask"] )
if self.padding_side == "right":
# Use `-1` since `0` in `global_attention_mask` means `local attention` instead of `not to attend`
a__ : Tuple = (
encoded_inputs["global_attention_mask"] + [-1] * difference
)
elif self.padding_side == "left":
a__ : Dict = [-1] * difference + encoded_inputs[
"global_attention_mask"
]
else:
raise ValueError("Invalid padding strategy:" + str(self.padding_side ) )
return encoded_inputs
| 37 |
import inspect
import unittest
from datasets import load_dataset
from packaging import version
from transformers import BeitConfig
from transformers.models.auto import get_values
from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
from transformers.utils import cached_property, is_torch_available, is_vision_available
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, _config_zero_init, floats_tensor, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from torch import nn
from transformers import (
MODEL_MAPPING,
BeitForImageClassification,
BeitForMaskedImageModeling,
BeitForSemanticSegmentation,
BeitModel,
)
from transformers.models.beit.modeling_beit import BEIT_PRETRAINED_MODEL_ARCHIVE_LIST
if is_vision_available():
import PIL
from PIL import Image
from transformers import BeitImageProcessor
class A__ :
"""simple docstring"""
def __init__( self : Optional[int] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Optional[int]=100 , lowerCamelCase__ : str=13 , lowerCamelCase__ : Optional[int]=30 , lowerCamelCase__ : Union[str, Any]=2 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : List[Any]=True , lowerCamelCase__ : Tuple=True , lowerCamelCase__ : int=32 , lowerCamelCase__ : Union[str, Any]=4 , lowerCamelCase__ : Dict=4 , lowerCamelCase__ : Union[str, Any]=37 , lowerCamelCase__ : List[Any]="gelu" , lowerCamelCase__ : List[Any]=0.1 , lowerCamelCase__ : int=0.1 , lowerCamelCase__ : Union[str, Any]=10 , lowerCamelCase__ : str=0.02 , lowerCamelCase__ : Tuple=3 , lowerCamelCase__ : Dict=None , lowerCamelCase__ : List[str]=[0, 1, 2, 3] , ):
a__ : Dict = parent
a__ : Dict = 100
a__ : Optional[int] = batch_size
a__ : Union[str, Any] = image_size
a__ : Any = patch_size
a__ : Optional[Any] = num_channels
a__ : int = is_training
a__ : List[str] = use_labels
a__ : Optional[Any] = hidden_size
a__ : List[Any] = num_hidden_layers
a__ : str = num_attention_heads
a__ : str = intermediate_size
a__ : int = hidden_act
a__ : List[Any] = hidden_dropout_prob
a__ : Dict = attention_probs_dropout_prob
a__ : Union[str, Any] = type_sequence_label_size
a__ : Optional[Any] = initializer_range
a__ : List[str] = scope
a__ : int = out_indices
a__ : List[str] = num_labels
# in BeiT, the seq length equals the number of patches + 1 (we add 1 for the [CLS] token)
a__ : Optional[int] = (image_size // patch_size) ** 2
a__ : Union[str, Any] = num_patches + 1
def _UpperCamelCase( self : int ):
a__ : Optional[int] = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size] )
a__ : Optional[Any] = None
a__ : Tuple = None
if self.use_labels:
a__ : str = ids_tensor([self.batch_size] , self.type_sequence_label_size )
a__ : Dict = ids_tensor([self.batch_size, self.image_size, self.image_size] , self.num_labels )
a__ : Optional[int] = self.get_config()
return config, pixel_values, labels, pixel_labels
def _UpperCamelCase( self : Tuple ):
return BeitConfig(
vocab_size=self.vocab_size , image_size=self.image_size , patch_size=self.patch_size , num_channels=self.num_channels , hidden_size=self.hidden_size , num_hidden_layers=self.num_hidden_layers , num_attention_heads=self.num_attention_heads , intermediate_size=self.intermediate_size , hidden_act=self.hidden_act , hidden_dropout_prob=self.hidden_dropout_prob , attention_probs_dropout_prob=self.attention_probs_dropout_prob , is_decoder=lowerCamelCase__ , initializer_range=self.initializer_range , out_indices=self.out_indices , )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : int , lowerCamelCase__ : Union[str, Any] , lowerCamelCase__ : int , lowerCamelCase__ : Any ):
a__ : str = BeitModel(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = model(lowerCamelCase__ )
self.parent.assertEqual(result.last_hidden_state.shape , (self.batch_size, self.seq_length, self.hidden_size) )
def _UpperCamelCase( self : Tuple , lowerCamelCase__ : List[str] , lowerCamelCase__ : Any , lowerCamelCase__ : List[str] , lowerCamelCase__ : Tuple ):
a__ : int = BeitForMaskedImageModeling(config=lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[Any] = model(lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.seq_length - 1, self.vocab_size) )
def _UpperCamelCase( self : str , lowerCamelCase__ : Any , lowerCamelCase__ : str , lowerCamelCase__ : Optional[int] , lowerCamelCase__ : Any ):
a__ : List[str] = self.type_sequence_label_size
a__ : Optional[Any] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
# test greyscale images
a__ : Optional[Any] = 1
a__ : List[str] = BeitForImageClassification(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : List[str] = floats_tensor([self.batch_size, 1, self.image_size, self.image_size] )
a__ : Union[str, Any] = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(result.logits.shape , (self.batch_size, self.type_sequence_label_size) )
def _UpperCamelCase( self : Any , lowerCamelCase__ : str , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Any , lowerCamelCase__ : Dict ):
a__ : int = self.num_labels
a__ : List[str] = BeitForSemanticSegmentation(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.eval()
a__ : Tuple = model(lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
a__ : str = model(lowerCamelCase__ , labels=lowerCamelCase__ )
self.parent.assertEqual(
result.logits.shape , (self.batch_size, self.num_labels, self.image_size * 2, self.image_size * 2) )
def _UpperCamelCase( self : Optional[int] ):
a__ : Any = self.prepare_config_and_inputs()
a__, a__, a__, a__ : Union[str, Any] = config_and_inputs
a__ : Dict = {"pixel_values": pixel_values}
return config, inputs_dict
@require_torch
class A__ ( A__ , A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = (
(BeitModel, BeitForImageClassification, BeitForMaskedImageModeling, BeitForSemanticSegmentation)
if is_torch_available()
else ()
)
_lowercase = (
{
'feature-extraction': BeitModel,
'image-classification': BeitForImageClassification,
'image-segmentation': BeitForSemanticSegmentation,
}
if is_torch_available()
else {}
)
_lowercase = False
_lowercase = False
_lowercase = False
def _UpperCamelCase( self : Any ):
a__ : int = BeitModelTester(self )
a__ : Optional[Any] = ConfigTester(self , config_class=lowerCamelCase__ , has_text_modality=lowerCamelCase__ , hidden_size=37 )
def _UpperCamelCase( self : List[Any] ):
self.config_tester.run_common_tests()
@unittest.skip(reason="BEiT does not use inputs_embeds" )
def _UpperCamelCase( self : str ):
pass
@require_torch_multi_gpu
@unittest.skip(reason="BEiT has some layers using `add_module` which doesn't work well with `nn.DataParallel`" )
def _UpperCamelCase( self : Dict ):
pass
def _UpperCamelCase( self : Optional[Any] ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : List[str] = model_class(lowerCamelCase__ )
self.assertIsInstance(model.get_input_embeddings() , (nn.Module) )
a__ : Optional[int] = model.get_output_embeddings()
self.assertTrue(x is None or isinstance(lowerCamelCase__ , nn.Linear ) )
def _UpperCamelCase( self : str ):
a__, a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
a__ : int = model_class(lowerCamelCase__ )
a__ : str = inspect.signature(model.forward )
# signature.parameters is an OrderedDict => so arg_names order is deterministic
a__ : Optional[int] = [*signature.parameters.keys()]
a__ : Any = ["pixel_values"]
self.assertListEqual(arg_names[:1] , lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Dict = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*lowerCamelCase__ )
def _UpperCamelCase( self : int ):
a__ : int = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_masked_lm(*lowerCamelCase__ )
def _UpperCamelCase( self : List[Any] ):
a__ : Optional[int] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_image_classification(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[int] ):
a__ : Union[str, Any] = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_semantic_segmentation(*lowerCamelCase__ )
def _UpperCamelCase( self : Optional[Any] ):
if not self.model_tester.is_training:
return
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : str = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]:
continue
a__ : List[str] = model_class(lowerCamelCase__ )
model.to(lowerCamelCase__ )
model.train()
a__ : Any = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : Tuple = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : Tuple ):
a__, a__ : Optional[Any] = self.model_tester.prepare_config_and_inputs_for_common()
if not self.model_tester.is_training:
return
a__ : List[Any] = False
a__ : List[str] = True
for model_class in self.all_model_classes:
# we don't test BeitForMaskedImageModeling
if (
model_class in [*get_values(lowerCamelCase__ ), BeitForMaskedImageModeling]
or not model_class.supports_gradient_checkpointing
):
continue
a__ : Optional[Any] = model_class(lowerCamelCase__ )
model.gradient_checkpointing_enable()
model.to(lowerCamelCase__ )
model.train()
a__ : Union[str, Any] = self._prepare_for_class(lowerCamelCase__ , lowerCamelCase__ , return_labels=lowerCamelCase__ )
a__ : int = model(**lowerCamelCase__ ).loss
loss.backward()
def _UpperCamelCase( self : List[str] ):
a__, a__ : Tuple = self.model_tester.prepare_config_and_inputs_for_common()
a__ : Dict = _config_zero_init(lowerCamelCase__ )
for model_class in self.all_model_classes:
a__ : str = model_class(config=lowerCamelCase__ )
for name, param in model.named_parameters():
# we skip lambda parameters as these require special initial values
# determined by config.layer_scale_init_value
if "lambda" in name:
continue
if param.requires_grad:
self.assertIn(
((param.data.mean() * 1E9).round() / 1E9).item() , [0.0, 1.0] , msg=f'''Parameter {name} of model {model_class} seems not properly initialized''' , )
@slow
def _UpperCamelCase( self : Optional[int] ):
for model_name in BEIT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
a__ : Tuple = BeitModel.from_pretrained(lowerCamelCase__ )
self.assertIsNotNone(lowerCamelCase__ )
def UpperCamelCase_ ( ) -> Any:
a__ : List[Any] = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png" )
return image
@require_torch
@require_vision
class A__ ( unittest.TestCase ):
"""simple docstring"""
@cached_property
def _UpperCamelCase( self : Optional[int] ):
return BeitImageProcessor.from_pretrained("microsoft/beit-base-patch16-224" ) if is_vision_available() else None
@slow
def _UpperCamelCase( self : str ):
a__ : int = BeitForMaskedImageModeling.from_pretrained("microsoft/beit-base-patch16-224-pt22k" ).to(lowerCamelCase__ )
a__ : Optional[Any] = self.default_image_processor
a__ : Dict = prepare_img()
a__ : Optional[int] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).pixel_values.to(lowerCamelCase__ )
# prepare bool_masked_pos
a__ : Optional[Any] = torch.ones((1, 196) , dtype=torch.bool ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Any = model(pixel_values=lowerCamelCase__ , bool_masked_pos=lowerCamelCase__ )
a__ : Tuple = outputs.logits
# verify the logits
a__ : List[str] = torch.Size((1, 196, 8_192) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[int] = torch.tensor(
[[-3.2437, 0.5072, -13.9174], [-3.2456, 0.4948, -13.9401], [-3.2033, 0.5121, -13.8550]] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[bool_masked_pos][:3, :3] , lowerCamelCase__ , atol=1E-2 ) )
@slow
def _UpperCamelCase( self : Dict ):
a__ : str = BeitForImageClassification.from_pretrained("microsoft/beit-base-patch16-224" ).to(lowerCamelCase__ )
a__ : int = self.default_image_processor
a__ : List[Any] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Union[str, Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Union[str, Any] = torch.Size((1, 1_000) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = torch.tensor([-1.2385, -1.0987, -1.0108] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Tuple = 281
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : Any ):
a__ : Dict = BeitForImageClassification.from_pretrained("microsoft/beit-large-patch16-224-pt22k-ft22k" ).to(
lowerCamelCase__ )
a__ : str = self.default_image_processor
a__ : List[str] = prepare_img()
a__ : Tuple = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Dict = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Optional[int] = torch.Size((1, 21_841) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : Optional[Any] = torch.tensor([1.6881, -0.2787, 0.5901] ).to(lowerCamelCase__ )
self.assertTrue(torch.allclose(logits[0, :3] , lowerCamelCase__ , atol=1E-4 ) )
a__ : Optional[Any] = 2_396
self.assertEqual(logits.argmax(-1 ).item() , lowerCamelCase__ )
@slow
def _UpperCamelCase( self : int ):
a__ : Optional[Any] = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : Tuple = model.to(lowerCamelCase__ )
a__ : List[Any] = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Tuple = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : Union[str, Any] = Image.open(ds[0]["file"] )
a__ : List[Any] = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : Optional[Any] = model(**lowerCamelCase__ )
a__ : List[str] = outputs.logits
# verify the logits
a__ : Tuple = torch.Size((1, 150, 160, 160) )
self.assertEqual(logits.shape , lowerCamelCase__ )
a__ : int = version.parse(PIL.__version__ ) < version.parse("9.0.0" )
if is_pillow_less_than_a:
a__ : Dict = torch.tensor(
[
[[-4.9225, -2.3954, -3.0522], [-2.8822, -1.0046, -1.7561], [-2.9549, -1.3228, -2.1347]],
[[-5.8168, -3.4129, -4.0778], [-3.8651, -2.2214, -3.0277], [-3.8356, -2.4643, -3.3535]],
[[-0.0078, 3.9952, 4.0754], [2.9856, 4.6944, 5.0035], [3.2413, 4.7813, 4.9969]],
] , device=lowerCamelCase__ , )
else:
a__ : Dict = torch.tensor(
[
[[-4.8960, -2.3688, -3.0355], [-2.8478, -0.9836, -1.7418], [-2.9449, -1.3332, -2.1456]],
[[-5.8081, -3.4124, -4.1006], [-3.8561, -2.2081, -3.0323], [-3.8365, -2.4601, -3.3669]],
[[-0.0309, 3.9868, 4.0540], [2.9640, 4.6877, 4.9976], [3.2081, 4.7690, 4.9942]],
] , device=lowerCamelCase__ , )
self.assertTrue(torch.allclose(logits[0, :3, :3, :3] , lowerCamelCase__ , atol=1E-4 ) )
@slow
def _UpperCamelCase( self : Tuple ):
a__ : str = BeitForSemanticSegmentation.from_pretrained("microsoft/beit-base-finetuned-ade-640-640" )
a__ : List[Any] = model.to(lowerCamelCase__ )
a__ : int = BeitImageProcessor(do_resize=lowerCamelCase__ , size=640 , do_center_crop=lowerCamelCase__ )
a__ : Optional[int] = load_dataset("hf-internal-testing/fixtures_ade20k" , split="test" )
a__ : str = Image.open(ds[0]["file"] )
a__ : str = image_processor(images=lowerCamelCase__ , return_tensors="pt" ).to(lowerCamelCase__ )
# forward pass
with torch.no_grad():
a__ : List[Any] = model(**lowerCamelCase__ )
a__ : Any = outputs.logits.detach().cpu()
a__ : List[Any] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ , target_sizes=[(500, 300)] )
a__ : Optional[int] = torch.Size((500, 300) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
a__ : List[str] = image_processor.post_process_semantic_segmentation(outputs=lowerCamelCase__ )
a__ : Any = torch.Size((160, 160) )
self.assertEqual(segmentation[0].shape , lowerCamelCase__ )
| 37 | 1 |
def UpperCamelCase_ ( ) -> int:
return [
a * b * (1_000 - a - b)
for a in range(1 , 999 )
for b in range(__a , 999 )
if (a * a + b * b == (1_000 - a - b) ** 2)
][0]
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 |
import re
import jax.numpy as jnp
from flax.traverse_util import flatten_dict, unflatten_dict
from jax.random import PRNGKey
from ..utils import logging
UpperCamelCase : Dict = logging.get_logger(__name__)
def UpperCamelCase_ ( __a ) -> Union[str, Any]:
a__ : Tuple = R"\w+[.]\d+"
a__ : List[Any] = re.findall(__a , __a )
for pat in pats:
a__ : Union[str, Any] = key.replace(__a , "_".join(pat.split("." ) ) )
return key
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : List[str] = pt_tuple_key[:-1] + ("scale",)
if (
any("norm" in str_ for str_ in pt_tuple_key )
and (pt_tuple_key[-1] == "bias")
and (pt_tuple_key[:-1] + ("bias",) not in random_flax_state_dict)
and (pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict)
):
a__ : Any = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
elif pt_tuple_key[-1] in ["weight", "gamma"] and pt_tuple_key[:-1] + ("scale",) in random_flax_state_dict:
a__ : Optional[Any] = pt_tuple_key[:-1] + ("scale",)
return renamed_pt_tuple_key, pt_tensor
# embedding
if pt_tuple_key[-1] == "weight" and pt_tuple_key[:-1] + ("embedding",) in random_flax_state_dict:
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("embedding",)
return renamed_pt_tuple_key, pt_tensor
# conv layer
a__ : List[str] = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight" and pt_tensor.ndim == 4:
a__ : str = pt_tensor.transpose(2 , 3 , 1 , 0 )
return renamed_pt_tuple_key, pt_tensor
# linear layer
a__ : Tuple = pt_tuple_key[:-1] + ("kernel",)
if pt_tuple_key[-1] == "weight":
a__ : Tuple = pt_tensor.T
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm weight
a__ : Optional[Any] = pt_tuple_key[:-1] + ("weight",)
if pt_tuple_key[-1] == "gamma":
return renamed_pt_tuple_key, pt_tensor
# old PyTorch layer norm bias
a__ : Union[str, Any] = pt_tuple_key[:-1] + ("bias",)
if pt_tuple_key[-1] == "beta":
return renamed_pt_tuple_key, pt_tensor
return pt_tuple_key, pt_tensor
def UpperCamelCase_ ( __a , __a , __a=42 ) -> str:
# Step 1: Convert pytorch tensor to numpy
a__ : Optional[int] = {k: v.numpy() for k, v in pt_state_dict.items()}
# Step 2: Since the model is stateless, get random Flax params
a__ : Tuple = flax_model.init_weights(PRNGKey(__a ) )
a__ : Optional[Any] = flatten_dict(__a )
a__ : Union[str, Any] = {}
# Need to change some parameters name to match Flax names
for pt_key, pt_tensor in pt_state_dict.items():
a__ : Optional[int] = rename_key(__a )
a__ : Optional[int] = tuple(renamed_pt_key.split("." ) )
# Correctly rename weight parameters
a__, a__ : Union[str, Any] = rename_key_and_reshape_tensor(__a , __a , __a )
if flax_key in random_flax_state_dict:
if flax_tensor.shape != random_flax_state_dict[flax_key].shape:
raise ValueError(
f'''PyTorch checkpoint seems to be incorrect. Weight {pt_key} was expected to be of shape '''
f'''{random_flax_state_dict[flax_key].shape}, but is {flax_tensor.shape}.''' )
# also add unexpected weight so that warning is thrown
a__ : str = jnp.asarray(__a )
return unflatten_dict(__a )
| 37 | 1 |
from transformers import HfArgumentParser, TensorFlowBenchmark, TensorFlowBenchmarkArguments
def UpperCamelCase_ ( ) -> int:
a__ : Any = HfArgumentParser(__a )
a__ : Any = parser.parse_args_into_dataclasses()[0]
a__ : Optional[int] = TensorFlowBenchmark(args=__a )
try:
a__ : Optional[int] = parser.parse_args_into_dataclasses()[0]
except ValueError as e:
a__ : Tuple = "Arg --no_{0} is no longer used, please use --no-{0} instead."
a__ : List[Any] = " ".join(str(__a ).split(" " )[:-1] )
a__ : str = ""
a__ : List[Any] = eval(str(__a ).split(" " )[-1] )
a__ : List[str] = []
for arg in depreciated_args:
# arg[2:] removes '--'
if arg[2:] in TensorFlowBenchmark.deprecated_args:
# arg[5:] removes '--no_'
full_error_msg += arg_error_msg.format(arg[5:] )
else:
wrong_args.append(__a )
if len(__a ) > 0:
a__ : Tuple = full_error_msg + begin_error_msg + str(__a )
raise ValueError(__a )
benchmark.run()
if __name__ == "__main__":
main()
| 37 |
from transformers import HfArgumentParser, TensorFlowBenchmark, TensorFlowBenchmarkArguments
def UpperCamelCase_ ( ) -> int:
a__ : Any = HfArgumentParser(__a )
a__ : Any = parser.parse_args_into_dataclasses()[0]
a__ : Optional[int] = TensorFlowBenchmark(args=__a )
try:
a__ : Optional[int] = parser.parse_args_into_dataclasses()[0]
except ValueError as e:
a__ : Tuple = "Arg --no_{0} is no longer used, please use --no-{0} instead."
a__ : List[Any] = " ".join(str(__a ).split(" " )[:-1] )
a__ : str = ""
a__ : List[Any] = eval(str(__a ).split(" " )[-1] )
a__ : List[str] = []
for arg in depreciated_args:
# arg[2:] removes '--'
if arg[2:] in TensorFlowBenchmark.deprecated_args:
# arg[5:] removes '--no_'
full_error_msg += arg_error_msg.format(arg[5:] )
else:
wrong_args.append(__a )
if len(__a ) > 0:
a__ : Tuple = full_error_msg + begin_error_msg + str(__a )
raise ValueError(__a )
benchmark.run()
if __name__ == "__main__":
main()
| 37 | 1 |
from itertools import permutations
def UpperCamelCase_ ( __a ) -> bool:
if num[3] % 2 != 0:
return False
if (num[2] + num[3] + num[4]) % 3 != 0:
return False
if num[5] % 5 != 0:
return False
a__ : Tuple = [7, 11, 13, 17]
for i, test in enumerate(__a ):
if (num[i + 4] * 100 + num[i + 5] * 10 + num[i + 6]) % test != 0:
return False
return True
def UpperCamelCase_ ( __a = 10 ) -> int:
return sum(
int("".join(map(__a , __a ) ) )
for num in permutations(range(__a ) )
if is_substring_divisible(__a ) )
if __name__ == "__main__":
print(f"""{solution() = }""")
| 37 |
import argparse
import ast
import logging
import os
import sys
import pandas as pd
import torch
from tqdm import tqdm
from transformers import BartForConditionalGeneration, RagRetriever, RagSequenceForGeneration, RagTokenForGeneration
from transformers import logging as transformers_logging
sys.path.append(os.path.join(os.getcwd())) # noqa: E402 # isort:skip
from utils_rag import exact_match_score, fa_score # noqa: E402 # isort:skip
UpperCamelCase : Optional[int] = logging.getLogger(__name__)
logging.basicConfig(level=logging.INFO)
transformers_logging.set_verbosity_info()
def UpperCamelCase_ ( __a ) -> Any:
if "token" in model_name_or_path:
return "rag_token"
if "sequence" in model_name_or_path:
return "rag_sequence"
if "bart" in model_name_or_path:
return "bart"
return None
def UpperCamelCase_ ( __a , __a , __a ) -> Any:
return max(metric_fn(__a , __a ) for gt in ground_truths )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = []
if args.gold_data_mode == "qa":
a__ : Any = pd.read_csv(__a , sep="\t" , header=__a )
for answer_list in data[1]:
a__ : Union[str, Any] = ast.literal_eval(__a )
answers.append(__a )
else:
a__ : List[str] = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : List[str] = [[reference] for reference in references]
a__ : List[str] = 0
for prediction, ground_truths in zip(__a , __a ):
total += 1
em += metric_max_over_ground_truths(__a , __a , __a )
fa += metric_max_over_ground_truths(__a , __a , __a )
a__ : Dict = 100.0 * em / total
a__ : Optional[Any] = 100.0 * fa / total
logger.info(f'''F1: {fa:.2f}''' )
logger.info(f'''EM: {em:.2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
a__ : Optional[Any] = args.k
a__ : str = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = [line.strip() for line in open(__a , "r" ).readlines()]
a__ : Tuple = 0
for hypo, reference in zip(__a , __a ):
a__ : Any = set(hypo.split("\t" )[:k] )
a__ : Union[str, Any] = set(reference.split("\t" ) )
total += 1
em += len(hypo_provenance & ref_provenance ) / k
a__ : Union[str, Any] = 100.0 * em / total
logger.info(f'''Precision@{k}: {em: .2f}''' )
def UpperCamelCase_ ( __a , __a , __a ) -> Optional[Any]:
def strip_title(__a ):
if title.startswith("\"" ):
a__ : Optional[Any] = title[1:]
if title.endswith("\"" ):
a__ : Union[str, Any] = title[:-1]
return title
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a , )["input_ids"].to(args.device )
a__ : Optional[int] = rag_model.rag.question_encoder(__a )
a__ : Union[str, Any] = question_enc_outputs[0]
a__ : Optional[int] = rag_model.retriever(
__a , question_enc_pool_output.cpu().detach().to(torch.floataa ).numpy() , prefix=rag_model.rag.generator.config.prefix , n_docs=rag_model.config.n_docs , return_tensors="pt" , )
a__ : List[Any] = rag_model.retriever.index.get_doc_dicts(result.doc_ids )
a__ : int = []
for docs in all_docs:
a__ : Optional[int] = [strip_title(__a ) for title in docs["title"]]
provenance_strings.append("\t".join(__a ) )
return provenance_strings
def UpperCamelCase_ ( __a , __a , __a ) -> Dict:
with torch.no_grad():
a__ : Optional[int] = rag_model.retriever.question_encoder_tokenizer.batch_encode_plus(
__a , return_tensors="pt" , padding=__a , truncation=__a )
a__ : Any = inputs_dict.input_ids.to(args.device )
a__ : Dict = inputs_dict.attention_mask.to(args.device )
a__ : Optional[int] = rag_model.generate( # rag_model overwrites generate
__a , attention_mask=__a , num_beams=args.num_beams , min_length=args.min_length , max_length=args.max_length , early_stopping=__a , num_return_sequences=1 , bad_words_ids=[[0, 0]] , )
a__ : int = rag_model.retriever.generator_tokenizer.batch_decode(__a , skip_special_tokens=__a )
if args.print_predictions:
for q, a in zip(__a , __a ):
logger.info("Q: {} - A: {}".format(__a , __a ) )
return answers
def UpperCamelCase_ ( ) -> List[str]:
a__ : int = argparse.ArgumentParser()
parser.add_argument(
"--model_type" , choices=["rag_sequence", "rag_token", "bart"] , type=__a , help=(
"RAG model type: rag_sequence, rag_token or bart, if none specified, the type is inferred from the"
" model_name_or_path"
) , )
parser.add_argument(
"--index_name" , default=__a , choices=["exact", "compressed", "legacy"] , type=__a , help="RAG model retriever type" , )
parser.add_argument(
"--index_path" , default=__a , type=__a , help="Path to the retrieval index" , )
parser.add_argument("--n_docs" , default=5 , type=__a , help="Number of retrieved docs" )
parser.add_argument(
"--model_name_or_path" , default=__a , type=__a , required=__a , help="Path to pretrained checkpoints or model identifier from huggingface.co/models" , )
parser.add_argument(
"--eval_mode" , choices=["e2e", "retrieval"] , default="e2e" , type=__a , help=(
"Evaluation mode, e2e calculates exact match and F1 of the downstream task, retrieval calculates"
" precision@k."
) , )
parser.add_argument("--k" , default=1 , type=__a , help="k for the precision@k calculation" )
parser.add_argument(
"--evaluation_set" , default=__a , type=__a , required=__a , help="Path to a file containing evaluation samples" , )
parser.add_argument(
"--gold_data_path" , default=__a , type=__a , required=__a , help="Path to a tab-separated file with gold samples" , )
parser.add_argument(
"--gold_data_mode" , default="qa" , type=__a , choices=["qa", "ans"] , help=(
"Format of the gold data file"
"qa - a single line in the following format: question [tab] answer_list"
"ans - a single line of the gold file contains the expected answer string"
) , )
parser.add_argument(
"--predictions_path" , type=__a , default="predictions.txt" , help="Name of the predictions file, to be stored in the checkpoints directory" , )
parser.add_argument(
"--eval_all_checkpoints" , action="store_true" , help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number" , )
parser.add_argument(
"--eval_batch_size" , default=8 , type=__a , help="Batch size per GPU/CPU for evaluation." , )
parser.add_argument(
"--recalculate" , help="Recalculate predictions even if the prediction file exists" , action="store_true" , )
parser.add_argument(
"--num_beams" , default=4 , type=__a , help="Number of beams to be used when generating answers" , )
parser.add_argument("--min_length" , default=1 , type=__a , help="Min length of the generated answers" )
parser.add_argument("--max_length" , default=50 , type=__a , help="Max length of the generated answers" )
parser.add_argument(
"--print_predictions" , action="store_true" , help="If True, prints predictions while evaluating." , )
parser.add_argument(
"--print_docs" , action="store_true" , help="If True, prints docs retried while generating." , )
a__ : int = parser.parse_args()
a__ : Dict = torch.device("cuda" if torch.cuda.is_available() else "cpu" )
return args
def UpperCamelCase_ ( __a ) -> Optional[int]:
a__ : Tuple = {}
if args.model_type is None:
a__ : List[str] = infer_model_type(args.model_name_or_path )
assert args.model_type is not None
if args.model_type.startswith("rag" ):
a__ : int = RagTokenForGeneration if args.model_type == "rag_token" else RagSequenceForGeneration
a__ : Tuple = args.n_docs
if args.index_name is not None:
a__ : Any = args.index_name
if args.index_path is not None:
a__ : int = args.index_path
else:
a__ : Optional[Any] = BartForConditionalGeneration
a__ : Tuple = (
[f.path for f in os.scandir(args.model_name_or_path ) if f.is_dir()]
if args.eval_all_checkpoints
else [args.model_name_or_path]
)
logger.info("Evaluate the following checkpoints: %s" , __a )
a__ : Any = get_scores if args.eval_mode == "e2e" else get_precision_at_k
a__ : Union[str, Any] = evaluate_batch_eae if args.eval_mode == "e2e" else evaluate_batch_retrieval
for checkpoint in checkpoints:
if os.path.exists(args.predictions_path ) and (not args.recalculate):
logger.info("Calculating metrics based on an existing predictions file: {}".format(args.predictions_path ) )
score_fn(__a , args.predictions_path , args.gold_data_path )
continue
logger.info("***** Running evaluation for {} *****".format(__a ) )
logger.info(" Batch size = %d" , args.eval_batch_size )
logger.info(" Predictions will be stored under {}".format(args.predictions_path ) )
if args.model_type.startswith("rag" ):
a__ : str = RagRetriever.from_pretrained(__a , **__a )
a__ : Optional[int] = model_class.from_pretrained(__a , retriever=__a , **__a )
model.retriever.init_retrieval()
else:
a__ : Dict = model_class.from_pretrained(__a , **__a )
model.to(args.device )
with open(args.evaluation_set , "r" ) as eval_file, open(args.predictions_path , "w" ) as preds_file:
a__ : List[Any] = []
for line in tqdm(__a ):
questions.append(line.strip() )
if len(__a ) == args.eval_batch_size:
a__ : Union[str, Any] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) + "\n" )
preds_file.flush()
a__ : Any = []
if len(__a ) > 0:
a__ : List[str] = evaluate_batch_fn(__a , __a , __a )
preds_file.write("\n".join(__a ) )
preds_file.flush()
score_fn(__a , args.predictions_path , args.gold_data_path )
if __name__ == "__main__":
UpperCamelCase : List[Any] = get_args()
main(args)
| 37 | 1 |
import math
def UpperCamelCase_ ( __a , __a ) -> Dict:
if 0 not in (x, y):
# We use the relation x^y = y*log10(x), where 10 is the base.
return y * math.logaa(__a )
else:
if x == 0: # 0 raised to any number is 0
return 0
elif y == 0:
return 1 # any number raised to 0 is 1
raise AssertionError("This should never happen" )
if __name__ == "__main__": # Main function
# Read two numbers from input and typecast them to int using map function.
# Here x is the base and y is the power.
UpperCamelCase : Tuple = """Enter the base and the power separated by a comma: """
UpperCamelCase , UpperCamelCase : str = map(int, input(prompt).split(""","""))
UpperCamelCase , UpperCamelCase : Tuple = map(int, input(prompt).split(""","""))
# We find the log of each number, using the function res(), which takes two
# arguments.
UpperCamelCase : Any = res(xa, ya)
UpperCamelCase : List[str] = res(xa, ya)
# We check for the largest number
if resa > resa:
print("""Largest number is""", xa, """^""", ya)
elif resa > resa:
print("""Largest number is""", xa, """^""", ya)
else:
print("""Both are equal""")
| 37 |
import logging
import os
from dataclasses import dataclass, field
from typing import Dict, Optional
import datasets
import numpy as np
import tensorflow as tf
from transformers import (
AutoConfig,
AutoTokenizer,
EvalPrediction,
HfArgumentParser,
PreTrainedTokenizer,
TFAutoModelForSequenceClassification,
TFTrainer,
TFTrainingArguments,
)
from transformers.utils import logging as hf_logging
hf_logging.set_verbosity_info()
hf_logging.enable_default_handler()
hf_logging.enable_explicit_format()
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a = None , ) -> str:
a__ : int = {}
if train_file is not None:
a__ : int = [train_file]
if eval_file is not None:
a__ : Union[str, Any] = [eval_file]
if test_file is not None:
a__ : str = [test_file]
a__ : Optional[Any] = datasets.load_dataset("csv" , data_files=__a )
a__ : List[Any] = list(ds[list(files.keys() )[0]].features.keys() )
a__ : str = features_name.pop(__a )
a__ : Dict = list(set(ds[list(files.keys() )[0]][label_name] ) )
a__ : str = {label: i for i, label in enumerate(__a )}
a__ : Tuple = tokenizer.model_input_names
a__ : List[str] = {}
if len(__a ) == 1:
for k in files.keys():
a__ : Optional[Any] = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
example[features_name[0]] , truncation=__a , max_length=__a , padding="max_length" ) , batched=__a , )
elif len(__a ) == 2:
for k in files.keys():
a__ : Dict = ds[k].map(
lambda __a : tokenizer.batch_encode_plus(
(example[features_name[0]], example[features_name[1]]) , truncation=__a , max_length=__a , padding="max_length" , ) , batched=__a , )
def gen_train():
for ex in transformed_ds[datasets.Split.TRAIN]:
a__ : str = {k: v for k, v in ex.items() if k in input_names}
a__ : str = labelaid[ex[label_name]]
yield (d, label)
def gen_val():
for ex in transformed_ds[datasets.Split.VALIDATION]:
a__ : Tuple = {k: v for k, v in ex.items() if k in input_names}
a__ : List[Any] = labelaid[ex[label_name]]
yield (d, label)
def gen_test():
for ex in transformed_ds[datasets.Split.TEST]:
a__ : List[Any] = {k: v for k, v in ex.items() if k in input_names}
a__ : Optional[int] = labelaid[ex[label_name]]
yield (d, label)
a__ : Optional[Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TRAIN in transformed_ds
else None
)
if train_ds is not None:
a__ : Optional[int] = train_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TRAIN] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.VALIDATION in transformed_ds
else None
)
if val_ds is not None:
a__ : Optional[Any] = val_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.VALIDATION] ) ) )
a__ : Union[str, Any] = (
tf.data.Dataset.from_generator(
__a , ({k: tf.intaa for k in input_names}, tf.intaa) , ({k: tf.TensorShape([None] ) for k in input_names}, tf.TensorShape([] )) , )
if datasets.Split.TEST in transformed_ds
else None
)
if test_ds is not None:
a__ : Tuple = test_ds.apply(tf.data.experimental.assert_cardinality(len(ds[datasets.Split.TEST] ) ) )
return train_ds, val_ds, test_ds, labelaid
UpperCamelCase : Optional[Any] = logging.getLogger(__name__)
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(metadata={'help': 'Which column contains the label'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the training file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the development file'} )
_lowercase = field(default=A__ , metadata={'help': 'The path of the test file'} )
_lowercase = field(
default=1_2_8 , metadata={
'help': (
'The maximum total input sequence length after tokenization. Sequences longer '
'than this will be truncated, sequences shorter will be padded.'
)
} , )
_lowercase = field(
default=A__ , metadata={'help': 'Overwrite the cached training and evaluation sets'} )
@dataclass
class A__ :
"""simple docstring"""
_lowercase = field(
metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained config name or path if not the same as model_name'} )
_lowercase = field(
default=A__ , metadata={'help': 'Pretrained tokenizer name or path if not the same as model_name'} )
_lowercase = field(default=A__ , metadata={'help': 'Set this flag to use fast tokenization.'} )
# If you want to tweak more attributes on your tokenizer, you should do it in a distinct script,
# or just modify its tokenizer_config.json.
_lowercase = field(
default=A__ , metadata={'help': 'Where do you want to store the pretrained models downloaded from huggingface.co'} , )
def UpperCamelCase_ ( ) -> Union[str, Any]:
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
a__ : str = HfArgumentParser((ModelArguments, DataTrainingArguments, TFTrainingArguments) )
a__, a__, a__ : str = parser.parse_args_into_dataclasses()
if (
os.path.exists(training_args.output_dir )
and os.listdir(training_args.output_dir )
and training_args.do_train
and not training_args.overwrite_output_dir
):
raise ValueError(
f'''Output directory ({training_args.output_dir}) already exists and is not empty. Use'''
" --overwrite_output_dir to overcome." )
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s" , datefmt="%m/%d/%Y %H:%M:%S" , level=logging.INFO , )
logger.info(
f'''n_replicas: {training_args.n_replicas}, distributed training: {bool(training_args.n_replicas > 1 )}, '''
f'''16-bits training: {training_args.fpaa}''' )
logger.info(f'''Training/evaluation parameters {training_args}''' )
# Load pretrained model and tokenizer
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
a__ : Union[str, Any] = AutoTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path , cache_dir=model_args.cache_dir , )
a__, a__, a__, a__ : Optional[Any] = get_tfds(
train_file=data_args.train_file , eval_file=data_args.dev_file , test_file=data_args.test_file , tokenizer=__a , label_column_id=data_args.label_column_id , max_seq_length=data_args.max_seq_length , )
a__ : Optional[int] = AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path , num_labels=len(__a ) , labelaid=__a , idalabel={id: label for label, id in labelaid.items()} , finetuning_task="text-classification" , cache_dir=model_args.cache_dir , )
with training_args.strategy.scope():
a__ : Any = TFAutoModelForSequenceClassification.from_pretrained(
model_args.model_name_or_path , from_pt=bool(".bin" in model_args.model_name_or_path ) , config=__a , cache_dir=model_args.cache_dir , )
def compute_metrics(__a ) -> Dict:
a__ : Union[str, Any] = np.argmax(p.predictions , axis=1 )
return {"acc": (preds == p.label_ids).mean()}
# Initialize our Trainer
a__ : Dict = TFTrainer(
model=__a , args=__a , train_dataset=__a , eval_dataset=__a , compute_metrics=__a , )
# Training
if training_args.do_train:
trainer.train()
trainer.save_model()
tokenizer.save_pretrained(training_args.output_dir )
# Evaluation
a__ : Optional[Any] = {}
if training_args.do_eval:
logger.info("*** Evaluate ***" )
a__ : Dict = trainer.evaluate()
a__ : int = os.path.join(training_args.output_dir , "eval_results.txt" )
with open(__a , "w" ) as writer:
logger.info("***** Eval results *****" )
for key, value in result.items():
logger.info(f''' {key} = {value}''' )
writer.write(f'''{key} = {value}\n''' )
results.update(__a )
return results
if __name__ == "__main__":
main()
| 37 | 1 |
import itertools
import random
import unittest
import numpy as np
from transformers import ASTFeatureExtractor
from transformers.testing_utils import require_torch, require_torchaudio
from transformers.utils.import_utils import is_torch_available
from ...test_sequence_feature_extraction_common import SequenceFeatureExtractionTestMixin
UpperCamelCase : Tuple = random.Random()
if is_torch_available():
import torch
def UpperCamelCase_ ( __a , __a=1.0 , __a=None , __a=None ) -> Dict:
if rng is None:
a__ : int = global_rng
a__ : Union[str, Any] = []
for batch_idx in range(shape[0] ):
values.append([] )
for _ in range(shape[1] ):
values[-1].append(rng.random() * scale )
return values
class A__ ( unittest.TestCase ):
"""simple docstring"""
def __init__( self : List[Any] , lowerCamelCase__ : List[Any] , lowerCamelCase__ : Tuple=7 , lowerCamelCase__ : Tuple=400 , lowerCamelCase__ : List[str]=2_000 , lowerCamelCase__ : Union[str, Any]=1 , lowerCamelCase__ : Optional[Any]=0.0 , lowerCamelCase__ : Any=16_000 , lowerCamelCase__ : str=True , lowerCamelCase__ : List[str]=True , ):
a__ : List[Any] = parent
a__ : Dict = batch_size
a__ : List[Any] = min_seq_length
a__ : Tuple = max_seq_length
a__ : Tuple = (self.max_seq_length - self.min_seq_length) // (self.batch_size - 1)
a__ : Dict = feature_size
a__ : Any = padding_value
a__ : List[Any] = sampling_rate
a__ : Tuple = return_attention_mask
a__ : int = do_normalize
def _UpperCamelCase( self : Tuple ):
return {
"feature_size": self.feature_size,
"padding_value": self.padding_value,
"sampling_rate": self.sampling_rate,
"return_attention_mask": self.return_attention_mask,
"do_normalize": self.do_normalize,
}
def _UpperCamelCase( self : str , lowerCamelCase__ : str=False , lowerCamelCase__ : Optional[Any]=False ):
def _flatten(lowerCamelCase__ : Dict ):
return list(itertools.chain(*lowerCamelCase__ ) )
if equal_length:
a__ : List[str] = floats_list((self.batch_size, self.max_seq_length) )
else:
# make sure that inputs increase in size
a__ : Tuple = [
_flatten(floats_list((x, self.feature_size) ) )
for x in range(self.min_seq_length , self.max_seq_length , self.seq_length_diff )
]
if numpify:
a__ : Any = [np.asarray(lowerCamelCase__ ) for x in speech_inputs]
return speech_inputs
@require_torch
@require_torchaudio
class A__ ( A__ , unittest.TestCase ):
"""simple docstring"""
_lowercase = ASTFeatureExtractor
def _UpperCamelCase( self : Dict ):
a__ : List[Any] = ASTFeatureExtractionTester(self )
def _UpperCamelCase( self : str ):
# Tests that all call wrap to encode_plus and batch_encode_plus
a__ : Optional[Any] = self.feature_extraction_class(**self.feat_extract_tester.prepare_feat_extract_dict() )
# create three inputs of length 800, 1000, and 1200
a__ : List[Any] = [floats_list((1, x) )[0] for x in range(800 , 1_400 , 200 )]
a__ : str = [np.asarray(lowerCamelCase__ ) for speech_input in speech_inputs]
# Test not batched input
a__ : Tuple = feat_extract(speech_inputs[0] , return_tensors="np" ).input_values
a__ : Union[str, Any] = feat_extract(np_speech_inputs[0] , return_tensors="np" ).input_values
self.assertTrue(np.allclose(lowerCamelCase__ , lowerCamelCase__ , atol=1E-3 ) )
# Test batched
a__ : List[str] = feat_extract(lowerCamelCase__ , padding=lowerCamelCase__ , return_tensors="np" ).input_values
a__ : str = feat_extract(lowerCamelCase__ , padding=lowerCamelCase__ , return_tensors="np" ).input_values
for enc_seq_a, enc_seq_a in zip(lowerCamelCase__ , lowerCamelCase__ ):
self.assertTrue(np.allclose(lowerCamelCase__ , lowerCamelCase__ , atol=1E-3 ) )
# Test 2-D numpy arrays are batched.
a__ : str = [floats_list((1, x) )[0] for x in (800, 800, 800)]
a__ : List[Any] = np.asarray(lowerCamelCase__ )
a__ : int = feat_extract(lowerCamelCase__ , return_tensors="np" ).input_values
a__ : str = feat_extract(lowerCamelCase__ , return_tensors="np" ).input_values
for enc_seq_a, enc_seq_a in zip(lowerCamelCase__ , lowerCamelCase__ ):
self.assertTrue(np.allclose(lowerCamelCase__ , lowerCamelCase__ , atol=1E-3 ) )
@require_torch
def _UpperCamelCase( self : str ):
import torch
a__ : str = self.feature_extraction_class(**self.feat_extract_tester.prepare_feat_extract_dict() )
a__ : int = np.random.rand(100 ).astype(np.floataa )
a__ : List[str] = np_speech_inputs.tolist()
for inputs in [py_speech_inputs, np_speech_inputs]:
a__ : Union[str, Any] = feature_extractor.pad([{"input_values": inputs}] , return_tensors="np" )
self.assertTrue(np_processed.input_values.dtype == np.floataa )
a__ : str = feature_extractor.pad([{"input_values": inputs}] , return_tensors="pt" )
self.assertTrue(pt_processed.input_values.dtype == torch.floataa )
def _UpperCamelCase( self : Dict , lowerCamelCase__ : List[str] ):
from datasets import load_dataset
a__ : Optional[Any] = load_dataset("hf-internal-testing/librispeech_asr_dummy" , "clean" , split="validation" )
# automatic decoding with librispeech
a__ : Union[str, Any] = ds.sort("id" ).select(range(lowerCamelCase__ ) )[:num_samples]["audio"]
return [x["array"] for x in speech_samples]
@require_torch
def _UpperCamelCase( self : str ):
# fmt: off
a__ : List[Any] = torch.tensor(
[-0.9894, -1.2776, -0.9066, -1.2776, -0.9349, -1.2609, -1.0386, -1.2776,
-1.1561, -1.2776, -1.2052, -1.2723, -1.2190, -1.2132, -1.2776, -1.1133,
-1.1953, -1.1343, -1.1584, -1.2203, -1.1770, -1.2474, -1.2381, -1.1936,
-0.9270, -0.8317, -0.8049, -0.7706, -0.7565, -0.7869] )
# fmt: on
a__ : Union[str, Any] = self._load_datasamples(1 )
a__ : Dict = ASTFeatureExtractor()
a__ : List[str] = feature_extractor(lowerCamelCase__ , return_tensors="pt" ).input_values
self.assertEquals(input_values.shape , (1, 1_024, 128) )
self.assertTrue(torch.allclose(input_values[0, 0, :30] , lowerCamelCase__ , atol=1E-4 ) )
| 37 |
import argparse
import collections
import json
import os
import re
import string
import sys
import numpy as np
UpperCamelCase : List[str] = re.compile(r"""\b(a|an|the)\b""", re.UNICODE)
UpperCamelCase : Union[str, Any] = None
def UpperCamelCase_ ( ) -> List[str]:
a__ : List[Any] = argparse.ArgumentParser("Official evaluation script for SQuAD version 2.0." )
parser.add_argument("data_file" , metavar="data.json" , help="Input data JSON file." )
parser.add_argument("pred_file" , metavar="pred.json" , help="Model predictions." )
parser.add_argument(
"--out-file" , "-o" , metavar="eval.json" , help="Write accuracy metrics to file (default is stdout)." )
parser.add_argument(
"--na-prob-file" , "-n" , metavar="na_prob.json" , help="Model estimates of probability of no answer." )
parser.add_argument(
"--na-prob-thresh" , "-t" , type=__a , default=1.0 , help="Predict \"\" if no-answer probability exceeds this (default = 1.0)." , )
parser.add_argument(
"--out-image-dir" , "-p" , metavar="out_images" , default=__a , help="Save precision-recall curves to directory." )
parser.add_argument("--verbose" , "-v" , action="store_true" )
if len(sys.argv ) == 1:
parser.print_help()
sys.exit(1 )
return parser.parse_args()
def UpperCamelCase_ ( __a ) -> str:
a__ : Optional[Any] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : Dict = bool(qa["answers"]["text"] )
return qid_to_has_ans
def UpperCamelCase_ ( __a ) -> List[Any]:
def remove_articles(__a ):
return ARTICLES_REGEX.sub(" " , __a )
def white_space_fix(__a ):
return " ".join(text.split() )
def remove_punc(__a ):
a__ : Union[str, Any] = set(string.punctuation )
return "".join(ch for ch in text if ch not in exclude )
def lower(__a ):
return text.lower()
return white_space_fix(remove_articles(remove_punc(lower(__a ) ) ) )
def UpperCamelCase_ ( __a ) -> Dict:
if not s:
return []
return normalize_answer(__a ).split()
def UpperCamelCase_ ( __a , __a ) -> str:
return int(normalize_answer(__a ) == normalize_answer(__a ) )
def UpperCamelCase_ ( __a , __a ) -> Dict:
a__ : int = get_tokens(__a )
a__ : Optional[Any] = get_tokens(__a )
a__ : Any = collections.Counter(__a ) & collections.Counter(__a )
a__ : Dict = sum(common.values() )
if len(__a ) == 0 or len(__a ) == 0:
# If either is no-answer, then F1 is 1 if they agree, 0 otherwise
return int(gold_toks == pred_toks )
if num_same == 0:
return 0
a__ : Tuple = 1.0 * num_same / len(__a )
a__ : str = 1.0 * num_same / len(__a )
a__ : str = (2 * precision * recall) / (precision + recall)
return fa
def UpperCamelCase_ ( __a , __a ) -> int:
a__ : List[str] = {}
a__ : Optional[int] = {}
for article in dataset:
for p in article["paragraphs"]:
for qa in p["qas"]:
a__ : List[Any] = qa["id"]
a__ : Dict = [t for t in qa["answers"]["text"] if normalize_answer(__a )]
if not gold_answers:
# For unanswerable questions, only correct answer is empty string
a__ : Tuple = [""]
if qid not in preds:
print(f'''Missing prediction for {qid}''' )
continue
a__ : Tuple = preds[qid]
# Take max over all gold answers
a__ : Optional[int] = max(compute_exact(__a , __a ) for a in gold_answers )
a__ : str = max(compute_fa(__a , __a ) for a in gold_answers )
return exact_scores, fa_scores
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
a__ : Optional[Any] = {}
for qid, s in scores.items():
a__ : Dict = na_probs[qid] > na_prob_thresh
if pred_na:
a__ : Dict = float(not qid_to_has_ans[qid] )
else:
a__ : Optional[Any] = s
return new_scores
def UpperCamelCase_ ( __a , __a , __a=None ) -> Tuple:
if not qid_list:
a__ : Union[str, Any] = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores.values() ) / total),
("f1", 100.0 * sum(fa_scores.values() ) / total),
("total", total),
] )
else:
a__ : int = len(__a )
return collections.OrderedDict(
[
("exact", 100.0 * sum(exact_scores[k] for k in qid_list ) / total),
("f1", 100.0 * sum(fa_scores[k] for k in qid_list ) / total),
("total", total),
] )
def UpperCamelCase_ ( __a , __a , __a ) -> List[str]:
for k in new_eval:
a__ : Optional[Any] = new_eval[k]
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[int]:
plt.step(__a , __a , color="b" , alpha=0.2 , where="post" )
plt.fill_between(__a , __a , step="post" , alpha=0.2 , color="b" )
plt.xlabel("Recall" )
plt.ylabel("Precision" )
plt.xlim([0.0, 1.05] )
plt.ylim([0.0, 1.05] )
plt.title(__a )
plt.savefig(__a )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a , __a=None , __a=None ) -> Dict:
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
a__ : Any = 0.0
a__ : Optional[int] = 1.0
a__ : Optional[int] = 0.0
a__ : Any = [1.0]
a__ : Tuple = [0.0]
a__ : List[str] = 0.0
for i, qid in enumerate(__a ):
if qid_to_has_ans[qid]:
true_pos += scores[qid]
a__ : Any = true_pos / float(i + 1 )
a__ : int = true_pos / float(__a )
if i == len(__a ) - 1 or na_probs[qid] != na_probs[qid_list[i + 1]]:
# i.e., if we can put a threshold after this point
avg_prec += cur_p * (cur_r - recalls[-1])
precisions.append(__a )
recalls.append(__a )
if out_image:
plot_pr_curve(__a , __a , __a , __a )
return {"ap": 100.0 * avg_prec}
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> str:
if out_image_dir and not os.path.exists(__a ):
os.makedirs(__a )
a__ : Optional[int] = sum(1 for v in qid_to_has_ans.values() if v )
if num_true_pos == 0:
return
a__ : Optional[int] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_exact.png" ) , title="Precision-Recall curve for Exact Match score" , )
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_f1.png" ) , title="Precision-Recall curve for F1 score" , )
a__ : str = {k: float(__a ) for k, v in qid_to_has_ans.items()}
a__ : Optional[Any] = make_precision_recall_eval(
__a , __a , __a , __a , out_image=os.path.join(__a , "pr_oracle.png" ) , title="Oracle Precision-Recall curve (binary task of HasAns vs. NoAns)" , )
merge_eval(__a , __a , "pr_exact" )
merge_eval(__a , __a , "pr_f1" )
merge_eval(__a , __a , "pr_oracle" )
def UpperCamelCase_ ( __a , __a , __a , __a ) -> str:
if not qid_list:
return
a__ : Optional[Any] = [na_probs[k] for k in qid_list]
a__ : str = np.ones_like(__a ) / float(len(__a ) )
plt.hist(__a , weights=__a , bins=20 , range=(0.0, 1.0) )
plt.xlabel("Model probability of no-answer" )
plt.ylabel("Proportion of dataset" )
plt.title(f'''Histogram of no-answer probability: {name}''' )
plt.savefig(os.path.join(__a , f'''na_prob_hist_{name}.png''' ) )
plt.clf()
def UpperCamelCase_ ( __a , __a , __a , __a ) -> Optional[Any]:
a__ : str = sum(1 for k in qid_to_has_ans if not qid_to_has_ans[k] )
a__ : Optional[Any] = num_no_ans
a__ : Dict = cur_score
a__ : Any = 0.0
a__ : Optional[Any] = sorted(__a , key=lambda __a : na_probs[k] )
for i, qid in enumerate(__a ):
if qid not in scores:
continue
if qid_to_has_ans[qid]:
a__ : Optional[int] = scores[qid]
else:
if preds[qid]:
a__ : str = -1
else:
a__ : Union[str, Any] = 0
cur_score += diff
if cur_score > best_score:
a__ : Any = cur_score
a__ : Dict = na_probs[qid]
return 100.0 * best_score / len(__a ), best_thresh
def UpperCamelCase_ ( __a , __a , __a , __a , __a , __a ) -> Any:
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__, a__ : Tuple = find_best_thresh(__a , __a , __a , __a )
a__ : Any = best_exact
a__ : Any = exact_thresh
a__ : List[Any] = best_fa
a__ : Optional[int] = fa_thresh
def UpperCamelCase_ ( ) -> Tuple:
with open(OPTS.data_file ) as f:
a__ : List[Any] = json.load(__a )
a__ : Any = dataset_json["data"]
with open(OPTS.pred_file ) as f:
a__ : int = json.load(__a )
if OPTS.na_prob_file:
with open(OPTS.na_prob_file ) as f:
a__ : List[str] = json.load(__a )
else:
a__ : Optional[int] = {k: 0.0 for k in preds}
a__ : Optional[Any] = make_qid_to_has_ans(__a ) # maps qid to True/False
a__ : List[Any] = [k for k, v in qid_to_has_ans.items() if v]
a__ : Union[str, Any] = [k for k, v in qid_to_has_ans.items() if not v]
a__, a__ : str = get_raw_scores(__a , __a )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : str = apply_no_ans_threshold(__a , __a , __a , OPTS.na_prob_thresh )
a__ : Tuple = make_eval_dict(__a , __a )
if has_ans_qids:
a__ : str = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "HasAns" )
if no_ans_qids:
a__ : List[Any] = make_eval_dict(__a , __a , qid_list=__a )
merge_eval(__a , __a , "NoAns" )
if OPTS.na_prob_file:
find_all_best_thresh(__a , __a , __a , __a , __a , __a )
if OPTS.na_prob_file and OPTS.out_image_dir:
run_precision_recall_analysis(__a , __a , __a , __a , __a , OPTS.out_image_dir )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "hasAns" )
histogram_na_prob(__a , __a , OPTS.out_image_dir , "noAns" )
if OPTS.out_file:
with open(OPTS.out_file , "w" ) as f:
json.dump(__a , __a )
else:
print(json.dumps(__a , indent=2 ) )
if __name__ == "__main__":
UpperCamelCase : Any = parse_args()
if OPTS.out_image_dir:
import matplotlib
matplotlib.use("""Agg""")
import matplotlib.pyplot as plt
main()
| 37 | 1 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.