Hugging Face's logo

Spaces:
apjanco
/
qwen2-vl-fmb-demo
Sleeping

App Files Community
qwen2-vl-fmb-demo
File size: 4,654 Bytes 
ec7a981
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aaf6a0c
 
ec7a981
 
 
d660402
 
ec7a981
 
aaf6a0c
ec7a981
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
 
import gradio as gr
import spaces
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info
import torch
from PIL import Image
import subprocess
from datetime import datetime
import numpy as np
import os


# subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)

# models = {
#     "Qwen/Qwen2-VL-2B-Instruct": AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-VL-2B-Instruct", trust_remote_code=True, torch_dtype="auto", _attn_implementation="flash_attention_2").cuda().eval()

# }
def array_to_image_path(image_array):
    # Convert numpy array to PIL Image
    img = Image.fromarray(np.uint8(image_array))
    
    # Generate a unique filename using timestamp
    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
    filename = f"image_{timestamp}.png"
    
    # Save the image
    img.save(filename)
    
    # Get the full path of the saved image
    full_path = os.path.abspath(filename)
    
    return full_path
    
models = {
    "qwen2-vl-fmb": Qwen2VLForConditionalGeneration.from_pretrained("fmb-quibdo/qwen2-vl-fmb", torch_dtype="auto").cuda().eval(),
    "qwen2-vl-fmb-7B": Qwen2VLForConditionalGeneration.from_pretrained("fmb-quibdo/fmb-qwen-vl-7b", torch_dtype="auto").cuda().eval(),
}

processors = {
    "qwen2-vl-fmb": AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct"),
    'qwen2-vl-fmb-7B': AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
}

DESCRIPTION = "[Qwen2-VL 2B or 7B Spanish Demo](https://huggingface.co/ajanco/qwen2-vl-fmb)"

kwargs = {}
kwargs['torch_dtype'] = torch.bfloat16

user_prompt = '<|user|>\n'
assistant_prompt = '<|assistant|>\n'
prompt_suffix = "<|end|>\n"

@spaces.GPU
def run_example(image, text_input=None, model_id="qwen2-vl-fmb"):
    image_path = array_to_image_path(image)
    
    print(image_path)
    model = models[model_id]
    processor = processors[model_id]

    prompt = f"{user_prompt}<|image_1|>\n{text_input}{prompt_suffix}{assistant_prompt}"
    image = Image.fromarray(image).convert("RGB")
    messages = [
    {
            "role": "user",
            "content": [
                {
                    "type": "image",
                    "image": image_path,
                },
                {"type": "text", "text": text_input},
            ],
        }
    ]
    
    # Preparation for inference
    text = processor.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    image_inputs, video_inputs = process_vision_info(messages)
    inputs = processor(
        text=[text],
        images=image_inputs,
        videos=video_inputs,
        padding=True,
        return_tensors="pt",
    )
    inputs = inputs.to("cuda")
    
    # Inference: Generation of the output
    generated_ids = model.generate(**inputs, max_new_tokens=4096)
    generated_ids_trimmed = [
        out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
    ]
    output_text = processor.batch_decode(
        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
    )
    
    return output_text[0]

css = """
  #output {
    height: 500px; 
    overflow: auto; 
    border: 1px solid #ccc; 
  }
"""

with gr.Blocks(css=css) as demo:
    gr.Markdown(DESCRIPTION)
    with gr.Tab(label="Qwen2-VL-7B Input"):
        with gr.Row():
            with gr.Column():
                input_img = gr.Image(label="Input Picture")
                model_selector = gr.Dropdown(choices=list(models.keys()), label="Model", value="qwen2-vl-fmb")
                text_input = gr.Textbox(label="Question", value="extract text in Spanish from the image")
                submit_btn = gr.Button(value="Submit")
            with gr.Column():
                output_text = gr.Textbox(label="Output Text")

        gr.Examples(
            examples=[
                ["assets/test0.png", "extract text in Spanish", text_input],
                ["assets/test1.png", "extract text in Spanish", text_input],
                ["assets/test2.png", "extract text in Spanish", text_input],
                ["assets/test3.png", "extract text in Spanish", text_input],
                ["assets/test4.png", "extract text in Spanish", text_input]
            ],
            inputs=[input_img, text_input],
            outputs=[output_text],
            fn=run_example,
            cache_examples=True,
            label="Try examples"
        )
        submit_btn.click(run_example, [input_img, text_input, model_selector], [output_text])

demo.queue(api_open=False)
demo.launch(debug=True)