Spaces:
Running
on
Zero
Running
on
Zero
gokaygokay
commited on
Commit
•
87966a5
1
Parent(s):
0c2f001
qwen2-VL
Browse files- app.py +58 -5
- requirements.txt +9 -2
app.py
CHANGED
@@ -9,8 +9,9 @@ from huggingface_hub import InferenceClient
|
|
9 |
import subprocess
|
10 |
import torch
|
11 |
from PIL import Image
|
12 |
-
from transformers import AutoProcessor, AutoModelForCausalLM
|
13 |
-
import
|
|
|
14 |
|
15 |
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
16 |
|
@@ -21,6 +22,10 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
|
|
21 |
florence_model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-large', trust_remote_code=True).to(device).eval()
|
22 |
florence_processor = AutoProcessor.from_pretrained('microsoft/Florence-2-large', trust_remote_code=True)
|
23 |
|
|
|
|
|
|
|
|
|
24 |
# Florence caption function
|
25 |
@spaces.GPU
|
26 |
def florence_caption(image):
|
@@ -44,6 +49,50 @@ def florence_caption(image):
|
|
44 |
)
|
45 |
return parsed_answer["<MORE_DETAILED_CAPTION>"]
|
46 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
47 |
# Load JSON files
|
48 |
def load_json_file(file_name):
|
49 |
file_path = os.path.join("data", file_name)
|
@@ -469,6 +518,7 @@ def create_interface():
|
|
469 |
with gr.Accordion("Image and Caption", open=False):
|
470 |
input_image = gr.Image(label="Input Image (optional)")
|
471 |
caption_output = gr.Textbox(label="Generated Caption", lines=3)
|
|
|
472 |
create_caption_button = gr.Button("Create Caption")
|
473 |
add_caption_button = gr.Button("Add Caption to Prompt")
|
474 |
|
@@ -488,14 +538,17 @@ def create_interface():
|
|
488 |
generate_text_button = gr.Button("Generate Prompt with LLM (Llama 3.1 70B)")
|
489 |
text_output = gr.Textbox(label="Generated Text", lines=10)
|
490 |
|
491 |
-
def create_caption(image):
|
492 |
if image is not None:
|
493 |
-
|
|
|
|
|
|
|
494 |
return ""
|
495 |
|
496 |
create_caption_button.click(
|
497 |
create_caption,
|
498 |
-
inputs=[input_image],
|
499 |
outputs=[caption_output]
|
500 |
)
|
501 |
|
|
|
9 |
import subprocess
|
10 |
import torch
|
11 |
from PIL import Image
|
12 |
+
from transformers import AutoProcessor, AutoModelForCausalLM, Qwen2VLForConditionalGeneration
|
13 |
+
from qwen_vl_utils import process_vision_info
|
14 |
+
import numpy as np
|
15 |
|
16 |
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
17 |
|
|
|
22 |
florence_model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-large', trust_remote_code=True).to(device).eval()
|
23 |
florence_processor = AutoProcessor.from_pretrained('microsoft/Florence-2-large', trust_remote_code=True)
|
24 |
|
25 |
+
# Initialize Qwen2-VL-2B model
|
26 |
+
qwen_model = Qwen2VLForConditionalGeneration.from_pretrained("Qwen/Qwen2-VL-2B-Instruct", trust_remote_code=True, torch_dtype="auto").to(device).eval()
|
27 |
+
qwen_processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct", trust_remote_code=True)
|
28 |
+
|
29 |
# Florence caption function
|
30 |
@spaces.GPU
|
31 |
def florence_caption(image):
|
|
|
49 |
)
|
50 |
return parsed_answer["<MORE_DETAILED_CAPTION>"]
|
51 |
|
52 |
+
# Qwen2-VL-2B caption function
|
53 |
+
@spaces.GPU
|
54 |
+
def qwen_caption(image):
|
55 |
+
if not isinstance(image, Image.Image):
|
56 |
+
image = Image.fromarray(image)
|
57 |
+
|
58 |
+
image_path = array_to_image_path(np.array(image))
|
59 |
+
|
60 |
+
messages = [
|
61 |
+
{
|
62 |
+
"role": "user",
|
63 |
+
"content": [
|
64 |
+
{
|
65 |
+
"type": "image",
|
66 |
+
"image": image_path,
|
67 |
+
},
|
68 |
+
{"type": "text", "text": "Describe this image in detail."},
|
69 |
+
],
|
70 |
+
}
|
71 |
+
]
|
72 |
+
|
73 |
+
text = qwen_processor.apply_chat_template(
|
74 |
+
messages, tokenize=False, add_generation_prompt=True
|
75 |
+
)
|
76 |
+
image_inputs, video_inputs = process_vision_info(messages)
|
77 |
+
inputs = qwen_processor(
|
78 |
+
text=[text],
|
79 |
+
images=image_inputs,
|
80 |
+
videos=video_inputs,
|
81 |
+
padding=True,
|
82 |
+
return_tensors="pt",
|
83 |
+
)
|
84 |
+
inputs = inputs.to(device)
|
85 |
+
|
86 |
+
generated_ids = qwen_model.generate(**inputs, max_new_tokens=256)
|
87 |
+
generated_ids_trimmed = [
|
88 |
+
out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
|
89 |
+
]
|
90 |
+
output_text = qwen_processor.batch_decode(
|
91 |
+
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
|
92 |
+
)
|
93 |
+
|
94 |
+
return output_text[0]
|
95 |
+
|
96 |
# Load JSON files
|
97 |
def load_json_file(file_name):
|
98 |
file_path = os.path.join("data", file_name)
|
|
|
518 |
with gr.Accordion("Image and Caption", open=False):
|
519 |
input_image = gr.Image(label="Input Image (optional)")
|
520 |
caption_output = gr.Textbox(label="Generated Caption", lines=3)
|
521 |
+
caption_model = gr.Radio(["Florence", "Qwen"], label="Caption Model", value="Florence")
|
522 |
create_caption_button = gr.Button("Create Caption")
|
523 |
add_caption_button = gr.Button("Add Caption to Prompt")
|
524 |
|
|
|
538 |
generate_text_button = gr.Button("Generate Prompt with LLM (Llama 3.1 70B)")
|
539 |
text_output = gr.Textbox(label="Generated Text", lines=10)
|
540 |
|
541 |
+
def create_caption(image, model):
|
542 |
if image is not None:
|
543 |
+
if model == "Florence":
|
544 |
+
return florence_caption(image)
|
545 |
+
elif model == "Qwen":
|
546 |
+
return qwen_caption(image)
|
547 |
return ""
|
548 |
|
549 |
create_caption_button.click(
|
550 |
create_caption,
|
551 |
+
inputs=[input_image, caption_model],
|
552 |
outputs=[caption_output]
|
553 |
)
|
554 |
|
requirements.txt
CHANGED
@@ -1,4 +1,11 @@
|
|
1 |
spaces
|
2 |
-
transformers
|
3 |
timm
|
4 |
-
openai==1.37.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
spaces
|
|
|
2 |
timm
|
3 |
+
openai==1.37.0
|
4 |
+
numpy==1.24.4
|
5 |
+
Pillow==10.3.0
|
6 |
+
Requests==2.31.0
|
7 |
+
torch
|
8 |
+
torchvision
|
9 |
+
git+https://github.com/huggingface/transformers.git
|
10 |
+
accelerate
|
11 |
+
qwen-vl-utils
|