Spaces:
Build error
Build error
Commit
·
4106f96
1
Parent(s):
b6646ba
fix openai compatability
Browse files
README.md
CHANGED
|
@@ -82,13 +82,15 @@ Optionally, you can set the following environment variables to customize the gen
|
|
| 82 |
- `MAX_NUM_ROWS`: The maximum number of rows to generate, defaults to `1000`.
|
| 83 |
- `DEFAULT_BATCH_SIZE`: The default batch size to use for generating the dataset, defaults to `5`.
|
| 84 |
|
| 85 |
-
Optionally, you can use different models and APIs.
|
| 86 |
|
| 87 |
-
- `BASE_URL`: The base URL for any OpenAI compatible API, e.g. `https://api-inference.huggingface.co/v1/`, `https://api.openai.com/v1/`.
|
| 88 |
-
- `MODEL`: The model to use for generating the dataset, e.g. `meta-llama/Meta-Llama-3.1-8B-Instruct`, `gpt-4o`.
|
| 89 |
- `API_KEY`: The API key to use for the generation API, e.g. `hf_...`, `sk-...`. If not provided, it will default to the provided `HF_TOKEN` environment variable.
|
| 90 |
-
- `MAGPIE_PRE_QUERY_TEMPLATE`: Enforce setting the pre-query template for Magpie. Llama3 and Qwen2 are supported out of the box and will use `"<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n"` and `"<|im_start|>user\n"` respectively. For other models, you can pass a custom pre-query template string.
|
| 91 |
|
|
|
|
|
|
|
|
|
|
| 92 |
|
| 93 |
Optionally, you can also push your datasets to Argilla for further curation by setting the following environment variables:
|
| 94 |
|
|
|
|
| 82 |
- `MAX_NUM_ROWS`: The maximum number of rows to generate, defaults to `1000`.
|
| 83 |
- `DEFAULT_BATCH_SIZE`: The default batch size to use for generating the dataset, defaults to `5`.
|
| 84 |
|
| 85 |
+
Optionally, you can use different models and APIs. For providers outside of Hugging Face, we provide an integration through [LiteLLM](https://docs.litellm.ai/docs/providers).
|
| 86 |
|
| 87 |
+
- `BASE_URL`: The base URL for any OpenAI compatible API, e.g. `https://api-inference.huggingface.co/v1/`, `https://api.openai.com/v1/`, `http://127.0.0.1:11434/v1/`.
|
| 88 |
+
- `MODEL`: The model to use for generating the dataset, e.g. `meta-llama/Meta-Llama-3.1-8B-Instruct`, `openai/gpt-4o`, `ollama/llama3.1`.
|
| 89 |
- `API_KEY`: The API key to use for the generation API, e.g. `hf_...`, `sk-...`. If not provided, it will default to the provided `HF_TOKEN` environment variable.
|
|
|
|
| 90 |
|
| 91 |
+
SFT and Chat Data generation is only supported with Hugging Face Inference Endpoints , and you can set the following environment variables use it with models other than Llama3 and Qwen2.
|
| 92 |
+
|
| 93 |
+
- `MAGPIE_PRE_QUERY_TEMPLATE`: Enforce setting the pre-query template for Magpie, which is only supported with Hugging Face Inference Endpoints. Llama3 and Qwen2 are supported out of the box and will use `"<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n"` and `"<|im_start|>user\n"` respectively. For other models, you can pass a custom pre-query template string.
|
| 94 |
|
| 95 |
Optionally, you can also push your datasets to Argilla for further curation by setting the following environment variables:
|
| 96 |
|
app.py
CHANGED
|
@@ -1,3 +1,8 @@
|
|
|
|
|
|
|
|
| 1 |
from synthetic_dataset_generator import launch
|
| 2 |
|
|
|
|
|
|
|
|
|
|
| 3 |
launch()
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
|
| 3 |
from synthetic_dataset_generator import launch
|
| 4 |
|
| 5 |
+
os.environ["BASE_URL"] = "http://localhost:11434/v1"
|
| 6 |
+
os.environ["MODEL"] = "llama3.1"
|
| 7 |
+
|
| 8 |
launch()
|
src/synthetic_dataset_generator/apps/chat.py
CHANGED
|
@@ -20,6 +20,7 @@ from synthetic_dataset_generator.apps.base import (
|
|
| 20 |
validate_push_to_hub,
|
| 21 |
)
|
| 22 |
from synthetic_dataset_generator.constants import (
|
|
|
|
| 23 |
DEFAULT_BATCH_SIZE,
|
| 24 |
MODEL,
|
| 25 |
SFT_AVAILABLE,
|
|
@@ -413,8 +414,8 @@ with gr.Blocks() as app:
|
|
| 413 |
[
|
| 414 |
"## Supervised Fine-Tuning not available",
|
| 415 |
"",
|
| 416 |
-
f"This tool relies on the [Magpie](https://arxiv.org/abs/2406.08464) prequery template, which is not implemented for the {MODEL}
|
| 417 |
-
"Use Llama3 or Qwen2 models
|
| 418 |
]
|
| 419 |
)
|
| 420 |
)
|
|
|
|
| 20 |
validate_push_to_hub,
|
| 21 |
)
|
| 22 |
from synthetic_dataset_generator.constants import (
|
| 23 |
+
BASE_URL,
|
| 24 |
DEFAULT_BATCH_SIZE,
|
| 25 |
MODEL,
|
| 26 |
SFT_AVAILABLE,
|
|
|
|
| 414 |
[
|
| 415 |
"## Supervised Fine-Tuning not available",
|
| 416 |
"",
|
| 417 |
+
f"This tool relies on the [Magpie](https://arxiv.org/abs/2406.08464) prequery template, which is not implemented for the {MODEL} with {BASE_URL}.",
|
| 418 |
+
"Use Llama3 or Qwen2 models with Hugging Face Inference Endpoints.",
|
| 419 |
]
|
| 420 |
)
|
| 421 |
)
|
src/synthetic_dataset_generator/constants.py
CHANGED
|
@@ -19,6 +19,8 @@ MAX_NUM_TOKENS = int(os.getenv("MAX_NUM_TOKENS", 2048))
|
|
| 19 |
MAX_NUM_ROWS: str | int = int(os.getenv("MAX_NUM_ROWS", 1000))
|
| 20 |
DEFAULT_BATCH_SIZE = int(os.getenv("DEFAULT_BATCH_SIZE", 5))
|
| 21 |
MODEL = os.getenv("MODEL", "meta-llama/Meta-Llama-3.1-8B-Instruct")
|
|
|
|
|
|
|
| 22 |
_API_KEY = os.getenv("API_KEY")
|
| 23 |
if _API_KEY:
|
| 24 |
API_KEYS = [_API_KEY]
|
|
@@ -27,12 +29,9 @@ else:
|
|
| 27 |
os.getenv(f"HF_TOKEN_{i}") for i in range(1, 10)
|
| 28 |
]
|
| 29 |
API_KEYS = [token for token in API_KEYS if token]
|
| 30 |
-
BASE_URL = os.getenv("BASE_URL", "https://api-inference.huggingface.co/v1/")
|
| 31 |
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
"API_KEY is not set. Ensure you have set the API_KEY environment variable that has access to the Hugging Face Inference Endpoints."
|
| 35 |
-
)
|
| 36 |
llama_options = ["llama3", "llama-3", "llama 3"]
|
| 37 |
qwen_options = ["qwen2", "qwen-2", "qwen 2"]
|
| 38 |
if os.getenv("MAGPIE_PRE_QUERY_TEMPLATE"):
|
|
@@ -54,14 +53,16 @@ elif MODEL.lower() in qwen_options or any(
|
|
| 54 |
):
|
| 55 |
SFT_AVAILABLE = True
|
| 56 |
MAGPIE_PRE_QUERY_TEMPLATE = "qwen2"
|
| 57 |
-
|
|
|
|
| 58 |
SFT_AVAILABLE = False
|
|
|
|
|
|
|
| 59 |
warnings.warn(
|
| 60 |
-
"`SFT_AVAILABLE` is set to `False
|
| 61 |
)
|
| 62 |
MAGPIE_PRE_QUERY_TEMPLATE = None
|
| 63 |
|
| 64 |
-
|
| 65 |
# Embeddings
|
| 66 |
STATIC_EMBEDDING_MODEL = "minishlab/potion-base-8M"
|
| 67 |
|
|
|
|
| 19 |
MAX_NUM_ROWS: str | int = int(os.getenv("MAX_NUM_ROWS", 1000))
|
| 20 |
DEFAULT_BATCH_SIZE = int(os.getenv("DEFAULT_BATCH_SIZE", 5))
|
| 21 |
MODEL = os.getenv("MODEL", "meta-llama/Meta-Llama-3.1-8B-Instruct")
|
| 22 |
+
BASE_URL = os.getenv("BASE_URL", default=None)
|
| 23 |
+
|
| 24 |
_API_KEY = os.getenv("API_KEY")
|
| 25 |
if _API_KEY:
|
| 26 |
API_KEYS = [_API_KEY]
|
|
|
|
| 29 |
os.getenv(f"HF_TOKEN_{i}") for i in range(1, 10)
|
| 30 |
]
|
| 31 |
API_KEYS = [token for token in API_KEYS if token]
|
|
|
|
| 32 |
|
| 33 |
+
# Determine if SFT is available
|
| 34 |
+
SFT_AVAILABLE = False
|
|
|
|
|
|
|
| 35 |
llama_options = ["llama3", "llama-3", "llama 3"]
|
| 36 |
qwen_options = ["qwen2", "qwen-2", "qwen 2"]
|
| 37 |
if os.getenv("MAGPIE_PRE_QUERY_TEMPLATE"):
|
|
|
|
| 53 |
):
|
| 54 |
SFT_AVAILABLE = True
|
| 55 |
MAGPIE_PRE_QUERY_TEMPLATE = "qwen2"
|
| 56 |
+
|
| 57 |
+
if BASE_URL:
|
| 58 |
SFT_AVAILABLE = False
|
| 59 |
+
|
| 60 |
+
if not SFT_AVAILABLE:
|
| 61 |
warnings.warn(
|
| 62 |
+
message="`SFT_AVAILABLE` is set to `False`. Use Hugging Face Inference Endpoints to generate chat data."
|
| 63 |
)
|
| 64 |
MAGPIE_PRE_QUERY_TEMPLATE = None
|
| 65 |
|
|
|
|
| 66 |
# Embeddings
|
| 67 |
STATIC_EMBEDDING_MODEL = "minishlab/potion-base-8M"
|
| 68 |
|
src/synthetic_dataset_generator/pipelines/textcat.py
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
import random
|
| 2 |
from typing import List
|
| 3 |
|
| 4 |
-
from distilabel.llms import InferenceEndpointsLLM
|
| 5 |
from distilabel.steps.tasks import (
|
| 6 |
GenerateTextClassificationData,
|
| 7 |
TextClassification,
|
|
@@ -61,39 +61,66 @@ class TextClassificationTask(BaseModel):
|
|
| 61 |
|
| 62 |
|
| 63 |
def get_prompt_generator():
|
| 64 |
-
|
| 65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
api_key=_get_next_api_key(),
|
| 67 |
model_id=MODEL,
|
| 68 |
base_url=BASE_URL,
|
| 69 |
-
structured_output=
|
| 70 |
-
generation_kwargs=
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
),
|
| 76 |
system_prompt=PROMPT_CREATION_PROMPT,
|
| 77 |
use_system_prompt=True,
|
| 78 |
)
|
|
|
|
| 79 |
prompt_generator.load()
|
| 80 |
return prompt_generator
|
| 81 |
|
| 82 |
|
| 83 |
def get_textcat_generator(difficulty, clarity, temperature, is_sample):
|
| 84 |
-
|
| 85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86 |
model_id=MODEL,
|
| 87 |
base_url=BASE_URL,
|
| 88 |
api_key=_get_next_api_key(),
|
| 89 |
-
generation_kwargs=
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
"top_p": 0.95,
|
| 95 |
-
},
|
| 96 |
-
),
|
| 97 |
difficulty=None if difficulty == "mixed" else difficulty,
|
| 98 |
clarity=None if clarity == "mixed" else clarity,
|
| 99 |
seed=random.randint(0, 2**32 - 1),
|
|
@@ -103,16 +130,28 @@ def get_textcat_generator(difficulty, clarity, temperature, is_sample):
|
|
| 103 |
|
| 104 |
|
| 105 |
def get_labeller_generator(system_prompt, labels, multi_label):
|
| 106 |
-
|
| 107 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 108 |
model_id=MODEL,
|
| 109 |
base_url=BASE_URL,
|
| 110 |
api_key=_get_next_api_key(),
|
| 111 |
-
generation_kwargs=
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
context=system_prompt,
|
| 117 |
available_labels=labels,
|
| 118 |
n=len(labels) if multi_label else 1,
|
|
|
|
| 1 |
import random
|
| 2 |
from typing import List
|
| 3 |
|
| 4 |
+
from distilabel.llms import InferenceEndpointsLLM, OpenAILLM
|
| 5 |
from distilabel.steps.tasks import (
|
| 6 |
GenerateTextClassificationData,
|
| 7 |
TextClassification,
|
|
|
|
| 61 |
|
| 62 |
|
| 63 |
def get_prompt_generator():
|
| 64 |
+
structured_output = {
|
| 65 |
+
"format": "json",
|
| 66 |
+
"schema": TextClassificationTask,
|
| 67 |
+
}
|
| 68 |
+
generation_kwargs = {
|
| 69 |
+
"temperature": 0.8,
|
| 70 |
+
"max_new_tokens": MAX_NUM_TOKENS,
|
| 71 |
+
}
|
| 72 |
+
if BASE_URL:
|
| 73 |
+
llm = OpenAILLM(
|
| 74 |
+
model=MODEL,
|
| 75 |
+
base_url=BASE_URL,
|
| 76 |
+
api_key=_get_next_api_key(),
|
| 77 |
+
structured_output=structured_output,
|
| 78 |
+
generation_kwargs=generation_kwargs,
|
| 79 |
+
)
|
| 80 |
+
else:
|
| 81 |
+
generation_kwargs["do_sample"] = True
|
| 82 |
+
llm = InferenceEndpointsLLM(
|
| 83 |
api_key=_get_next_api_key(),
|
| 84 |
model_id=MODEL,
|
| 85 |
base_url=BASE_URL,
|
| 86 |
+
structured_output=structured_output,
|
| 87 |
+
generation_kwargs=generation_kwargs,
|
| 88 |
+
)
|
| 89 |
+
|
| 90 |
+
prompt_generator = TextGeneration(
|
| 91 |
+
llm=llm,
|
|
|
|
| 92 |
system_prompt=PROMPT_CREATION_PROMPT,
|
| 93 |
use_system_prompt=True,
|
| 94 |
)
|
| 95 |
+
|
| 96 |
prompt_generator.load()
|
| 97 |
return prompt_generator
|
| 98 |
|
| 99 |
|
| 100 |
def get_textcat_generator(difficulty, clarity, temperature, is_sample):
|
| 101 |
+
generation_kwargs = {
|
| 102 |
+
"temperature": temperature,
|
| 103 |
+
"max_new_tokens": 256 if is_sample else MAX_NUM_TOKENS,
|
| 104 |
+
"top_p": 0.95,
|
| 105 |
+
}
|
| 106 |
+
if BASE_URL:
|
| 107 |
+
llm = OpenAILLM(
|
| 108 |
+
model=MODEL,
|
| 109 |
+
base_url=BASE_URL,
|
| 110 |
+
api_key=_get_next_api_key(),
|
| 111 |
+
generation_kwargs=generation_kwargs,
|
| 112 |
+
)
|
| 113 |
+
else:
|
| 114 |
+
generation_kwargs["do_sample"] = True
|
| 115 |
+
llm = InferenceEndpointsLLM(
|
| 116 |
model_id=MODEL,
|
| 117 |
base_url=BASE_URL,
|
| 118 |
api_key=_get_next_api_key(),
|
| 119 |
+
generation_kwargs=generation_kwargs,
|
| 120 |
+
)
|
| 121 |
+
|
| 122 |
+
textcat_generator = GenerateTextClassificationData(
|
| 123 |
+
llm=llm,
|
|
|
|
|
|
|
|
|
|
| 124 |
difficulty=None if difficulty == "mixed" else difficulty,
|
| 125 |
clarity=None if clarity == "mixed" else clarity,
|
| 126 |
seed=random.randint(0, 2**32 - 1),
|
|
|
|
| 130 |
|
| 131 |
|
| 132 |
def get_labeller_generator(system_prompt, labels, multi_label):
|
| 133 |
+
generation_kwargs = {
|
| 134 |
+
"temperature": 0.01,
|
| 135 |
+
"max_new_tokens": MAX_NUM_TOKENS,
|
| 136 |
+
}
|
| 137 |
+
|
| 138 |
+
if BASE_URL:
|
| 139 |
+
llm = OpenAILLM(
|
| 140 |
+
model=MODEL,
|
| 141 |
+
base_url=BASE_URL,
|
| 142 |
+
api_key=_get_next_api_key(),
|
| 143 |
+
generation_kwargs=generation_kwargs,
|
| 144 |
+
)
|
| 145 |
+
else:
|
| 146 |
+
llm = InferenceEndpointsLLM(
|
| 147 |
model_id=MODEL,
|
| 148 |
base_url=BASE_URL,
|
| 149 |
api_key=_get_next_api_key(),
|
| 150 |
+
generation_kwargs=generation_kwargs,
|
| 151 |
+
)
|
| 152 |
+
|
| 153 |
+
labeller_generator = TextClassification(
|
| 154 |
+
llm=llm,
|
| 155 |
context=system_prompt,
|
| 156 |
available_labels=labels,
|
| 157 |
n=len(labels) if multi_label else 1,
|