app.py

import gradio as gr
from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor, TextIteratorStreamer
from threading import Thread
from qwen_vl_utils import process_vision_info
import torch
import time

local_path = "Fancy-MLLM/R1-OneVision-7B"

model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    local_path, torch_dtype="auto", device_map="cpu"
)
processor = AutoProcessor.from_pretrained(local_path)

def generate_output(image, text, button_click):
    # Prepare input data
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "image", "image": image, 'min_pixels': 1003520, 'max_pixels': 12845056},
                {"type": "text", "text": text},
            ],
        }
    ]
    
    # Prepare inputs for the model
    text_input = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    # print(text_input)
    # import pdb; pdb.set_trace()
    image_inputs, video_inputs = process_vision_info(messages)
    inputs = processor(
        text=[text_input],
        images=image_inputs,
        videos=video_inputs,
        padding=True,
        return_tensors="pt",
    )
    inputs = inputs.to(model.device)

    streamer = TextIteratorStreamer(processor, skip_prompt=True, skip_special_tokens=True)
    generation_kwargs = dict(
        **inputs,
        streamer=streamer,
        max_new_tokens=4096,
        top_p=0.001,
        top_k=1,
        temperature=0.01,
        repetition_penalty=1.0,
    )
    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()
    generated_text = ''
    
    try:
        for new_text in streamer:
            generated_text += new_text
            yield f"‎{generated_text}"
            # print(f"Current text: {generated_text}")  # 调试输出
            # yield generated_text  # 直接输出原始文本
    except Exception as e:
        print(f"Error: {e}")
        yield f"Error occurred: {str(e)}"

Css = """
#output-markdown {
    overflow-y: auto;
    white-space: pre-wrap; 
    word-wrap: break-word;
}

#output-markdown .math {
    overflow-x: auto;
    max-width: 100%;
}
.markdown-text {
    white-space: pre-wrap;
    word-wrap: break-word;
}
#qwen-md .katex-display { display: inline; }
#qwen-md .katex-display>.katex { display: inline; }
#qwen-md .katex-display>.katex>.katex-html { display: inline; }
"""

# UI 组件
with gr.Blocks(css=Css) as demo:
    gr.HTML("""<center><font size=8>🦖 R1-OneVision Demo</center>""")

    with gr.Row():
        with gr.Column():
            input_image = gr.Image(type="pil", label="Upload"),
            input_text = gr.Textbox(label="input your question")
            with gr.Row():
                with gr.Column():
                    clear_btn = gr.ClearButton([*input_image, input_text])
                with gr.Column():
                    submit_btn = gr.Button("Submit", variant="primary")

        with gr.Column():
            output_text = gr.Markdown(
                label="Generated Response",
                max_height="80vh",
                min_height="50vh",
                container=True,
                latex_delimiters=[{
                                        "left": "\\(",
                                        "right": "\\)",
                                        "display": True
                                    }, {
                                        "left": "\\begin\{equation\}",
                                        "right": "\\end\{equation\}",
                                        "display": True
                                    }, {
                                        "left": "\\begin\{align\}",
                                        "right": "\\end\{align\}",
                                        "display": True
                                    }, {
                                        "left": "\\begin\{alignat\}",
                                        "right": "\\end\{alignat\}",
                                        "display": True
                                    }, {
                                        "left": "\\begin\{gather\}",
                                        "right": "\\end\{gather\}",
                                        "display": True
                                    }, {
                                        "left": "\\begin\{CD\}",
                                        "right": "\\end\{CD\}",
                                        "display": True
                                    }, {
                                        "left": "\\[",
                                        "right": "\\]",
                                        "display": True
                                    }],
                elem_id="qwen-md")
            


    submit_btn.click(
        fn=generate_output,
        inputs=[*input_image, input_text],
        outputs=output_text,
        queue=True
    )
demo.launch(share=True)