app.py

import gradio as gr
import requests
from PIL import Image
import torch
from transformers import AutoModel, AutoProcessor
import spaces

model_path = "YannQi/R-4B"

model = AutoModel.from_pretrained(
    model_path,
    torch_dtype=torch.float32,
    trust_remote_code=True,
).to("cuda")

processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True)

@spaces.GPU(duration=120)
def generate_response(message, history, thinking_mode):
    if not message:
        return "", history

    messages = []
    all_images = []

    for user_msg, asst_msg in history:
        # Process user message
        if isinstance(user_msg, str):
            user_content = [{"type": "text", "text": user_msg}]
        else:
            text = user_msg.get('text', '')
            files = user_msg.get('files', [])
            file_paths = [f.get('path', str(f)) for f in files]
            user_content = []
            img_paths = file_paths if isinstance(file_paths, list) else []
            for path in img_paths:
                try:
                    img = Image.open(path)
                    all_images.append(img)
                    user_content.append({"type": "image", "image": path})
                except:
                    pass
            if text:
                user_content.append({"type": "text", "text": text})
        messages.append({"role": "user", "content": user_content})

        # Process assistant message
        asst_text = asst_msg if isinstance(asst_msg, str) else asst_msg.get('text', '')
        messages.append({"role": "assistant", "content": [{"type": "text", "text": asst_text}]})

    # Current user message
    if isinstance(message, str):
        curr_text = message
        curr_files = []
    else:
        curr_text = message.get('text', '')
        curr_files = message.get('files', [])
    curr_user_content = []
    curr_images = []
    curr_file_paths = [f.get('path', str(f)) for f in curr_files]
    for path in curr_file_paths:
        if path.lower().endswith(('.png', '.jpg', '.jpeg', '.gif', '.bmp')):
            try:
                img = Image.open(path)
                curr_images.append(img)
                curr_user_content.append({"type": "image", "image": path})
            except:
                pass
    if curr_text:
        curr_user_content.append({"type": "text", "text": curr_text})
    if not curr_user_content:
        return "", history
    messages.append({"role": "user", "content": curr_user_content})

    # Apply chat template
    text = processor.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True,
        thinking_mode=thinking_mode
    )

    # All images
    all_images += curr_images

    # Process inputs
    inputs = processor(
        images=all_images if all_images else None,
        text=text,
        return_tensors="pt"
    ).to("cuda")

    # Generate
    with torch.no_grad():
        generated_ids = model.generate(
            **inputs,
            max_new_tokens=512,
            do_sample=True,
            temperature=0.7
        )
    output_ids = generated_ids[0][len(inputs.input_ids[0]):]
    output_text = processor.decode(
        output_ids,
        skip_special_tokens=True,
        clean_up_tokenization_spaces=False
    )

    # Prepare display for current user message
    user_display = message
    new_history = history + [(user_display, output_text)]

    return "", new_history

with gr.Blocks(title="Transformers Chat") as demo:
    gr.Markdown("# Using 🤗 Transformers to Chat")
    gr.Markdown("Select thinking mode: auto (auto-thinking), long (thinking), short (non-thinking). Default: auto.")
    chatbot = gr.Chatbot(type="tuples", height=500, label="Chat")
    with gr.Row():
        msg = gr.MultimodalTextbox(
            placeholder="Type your message or upload images...",
            file_types=[".jpg", ".jpeg", ".png", ".gif", ".bmp"],
            file_count="multiple",
            label="Message"
        )
        mode = gr.Dropdown(
            choices=["auto", "long", "short"],
            value="auto",
            label="Thinking Mode",
            interactive=True
        )
    with gr.Row():
        submit_btn = gr.Button("Send", variant="primary", scale=3)
        clear_btn = gr.Button("Clear", scale=1)
    submit_btn.click(generate_response, [msg, chatbot, mode], [msg, chatbot])
    msg.submit(generate_response, [msg, chatbot, mode], [msg, chatbot])
    clear_btn.click(lambda: ([], ""), None, [chatbot, msg], queue=False)

if __name__ == "__main__":
    demo.launch()