Fixed missing output for last prediction
Browse files
app.py
CHANGED
@@ -7,6 +7,8 @@ from transformers import BitsAndBytesConfig
|
|
7 |
|
8 |
from sentence_transformers import SentenceTransformer, util
|
9 |
|
|
|
|
|
10 |
quantization_config = BitsAndBytesConfig(
|
11 |
load_in_4bit=True,
|
12 |
bnb_4bit_compute_dtype=torch.float16
|
@@ -20,7 +22,8 @@ model = LlavaForConditionalGeneration.from_pretrained(
|
|
20 |
quantization_config=quantization_config,
|
21 |
device_map="auto",
|
22 |
# use_flash_attention_2=True,
|
23 |
-
low_cpu_mem_usage=True
|
|
|
24 |
)
|
25 |
|
26 |
MAXIMUM_PIXEL_VALUES = 3725568
|
@@ -63,6 +66,14 @@ def text_to_image(image, prompt, duplications: float):
|
|
63 |
batch = dict(input_ids=list(), attention_mask=list(), pixel_values=list())
|
64 |
else:
|
65 |
i += 1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
else:
|
67 |
batched_inputs.append(inputs)
|
68 |
|
@@ -73,8 +84,8 @@ def text_to_image(image, prompt, duplications: float):
|
|
73 |
batch['input_ids'] = batch['input_ids'].to(model.device)
|
74 |
batch['attention_mask'] = batch['attention_mask'].to(model.device)
|
75 |
batch['pixel_values'] = batch['pixel_values'].to(model.device)
|
76 |
-
output = model.generate(**batch, max_new_tokens=500, temperature=0.3)
|
77 |
-
|
78 |
# Unload GPU
|
79 |
batch['input_ids'].to('cpu')
|
80 |
batch['attention_mask'].to('cpu')
|
|
|
7 |
|
8 |
from sentence_transformers import SentenceTransformer, util
|
9 |
|
10 |
+
from transformers import PretrainedConfig
|
11 |
+
|
12 |
quantization_config = BitsAndBytesConfig(
|
13 |
load_in_4bit=True,
|
14 |
bnb_4bit_compute_dtype=torch.float16
|
|
|
22 |
quantization_config=quantization_config,
|
23 |
device_map="auto",
|
24 |
# use_flash_attention_2=True,
|
25 |
+
low_cpu_mem_usage=True,
|
26 |
+
# config=PretrainedConfig(do_sample=True)
|
27 |
)
|
28 |
|
29 |
MAXIMUM_PIXEL_VALUES = 3725568
|
|
|
66 |
batch = dict(input_ids=list(), attention_mask=list(), pixel_values=list())
|
67 |
else:
|
68 |
i += 1
|
69 |
+
if i >= len(inputs['pixel_values']) and len(batch['input_ids']) > 0:
|
70 |
+
batch['input_ids'] = torch.stack(batch['input_ids'], dim=0)
|
71 |
+
batch['attention_mask'] = torch.stack(batch['attention_mask'], dim=0)
|
72 |
+
batch['pixel_values'] = torch.stack(batch['pixel_values'], dim=0)
|
73 |
+
|
74 |
+
# Add to the batched_inputs
|
75 |
+
batched_inputs.append(batch)
|
76 |
+
batch = dict(input_ids=list(), attention_mask=list(), pixel_values=list())
|
77 |
else:
|
78 |
batched_inputs.append(inputs)
|
79 |
|
|
|
84 |
batch['input_ids'] = batch['input_ids'].to(model.device)
|
85 |
batch['attention_mask'] = batch['attention_mask'].to(model.device)
|
86 |
batch['pixel_values'] = batch['pixel_values'].to(model.device)
|
87 |
+
# output = model.generate(**batch, max_new_tokens=500, temperature=0.3)
|
88 |
+
output = model.generate(**batch, max_new_tokens=500)
|
89 |
# Unload GPU
|
90 |
batch['input_ids'].to('cpu')
|
91 |
batch['attention_mask'].to('cpu')
|