app.py

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
import gradio as gr
import threading
import time

# Global variables to store the model and tokenizer
model = None
tokenizer = None
model_loading_lock = threading.Lock()
model_loaded = False  # Status flag to indicate if the model is loaded

def load_model(model_name="deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"):
    global model, tokenizer, model_loaded
    with model_loading_lock:
        if not model_loaded:
            print("Loading model...")
            tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
            model = AutoModelForCausalLM.from_pretrained(
                model_name,
                device_map="sequential",
                torch_dtype=torch.float16,
                trust_remote_code=True,
                low_cpu_mem_usage=True,
                offload_folder="offload"
            )
            model_loaded = True
            print("Model loaded successfully.")
        else:
            print("Model already loaded.")

def check_model_status():
    """Check if the model is loaded and reload if necessary."""
    global model_loaded
    if not model_loaded:
        print("Model not loaded. Reloading...")
        load_model()
    return model_loaded

def chat(message, history, temperature, max_new_tokens):
    global model, tokenizer
    stop_tokens = ["|im_end|"]
    
    # Ensure the model is loaded before proceeding
    if not check_model_status():
        yield "Model is not ready. Please try again later.", ""
        return
    
    prompt = f"Human: {message}\n\nAssistant:"
    
    # Tokenize the input
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    
    # Stream the response
    start_time = time.time()
    token_count = 0
    
    # Create a TextStreamer for token streaming
    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
    
    generate_kwargs = dict(
        input_ids=inputs.input_ids,
        max_new_tokens=max_new_tokens,
        temperature=temperature,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id,
        streamer=streamer  # Use the TextStreamer here
    )
    
    # Create and start the thread with the model.generate function
    t = threading.Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()
    
    outputs = []
    for new_token in streamer:
        outputs.append(new_token)
        token_count += 1
        
        # Calculate tokens per second
        elapsed_time = time.time() - start_time
        tokens_per_second = token_count / elapsed_time if elapsed_time > 0 else 0
        
        # Update the token status
        token_status_value = f"Tokens Generated: {token_count}, Tokens/Second: {tokens_per_second:.2f}"
        yield "".join(outputs), token_status_value
        
        if any(stop_token in new_token for stop_token in stop_tokens):
            break

def reload_model_button():
    """Reload the model manually via a button."""
    global model_loaded
    model_loaded = False
    load_model()
    return "Model reloaded successfully."

# Function to periodically update the status text
def update_status_periodically(status_text):
    while True:
        time.sleep(5)  # Update every 5 seconds
        status = "Model is loaded and ready." if model_loaded else "Model is not loaded."
        status_text.value = status  # Update the value directly

# Gradio Interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("# DeepSeek-R1 Chatbot")
    gr.Markdown("DeepSeek-R1-Distill-Qwen-1.5B 모델을 사용한 대화 테스트용 데모입니다.")
    
    with gr.Row():
        chatbot = gr.Chatbot(height=600)
        textbox = gr.Textbox(placeholder="Enter your message...", container=False, scale=7)
    
    with gr.Row():
        send_button = gr.Button("Send")
        clear_button = gr.Button("Clear")
        reload_button = gr.Button("Reload Model")
    
    with gr.Row():
        temperature_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.1, label="Temperature")
        max_tokens_slider = gr.Slider(minimum=32, maximum=2048, value=2048, step=32, label="Max New Tokens")
    
    status_text = gr.Textbox(label="Model Status", value="Model not loaded yet.", interactive=False)
    token_status = gr.Textbox(label="Token Generation Status", value="", interactive=False)
    
    def respond(message, chat_history, temperature, max_new_tokens):
        bot_message = ""
        for partial_response, token_status_value in chat(message, chat_history, temperature, max_new_tokens):
            bot_message = partial_response
            yield "", chat_history + [(message, bot_message)], gr.update(value=token_status_value)
    
    send_button.click(respond, inputs=[textbox, chatbot, temperature_slider, max_tokens_slider], outputs=[textbox, chatbot, token_status])
    clear_button.click(lambda: [], None, chatbot)
    reload_button.click(reload_model_button, None, status_text)
    
    # Start a background thread to update the status text periodically
    threading.Thread(target=update_status_periodically, args=(status_text,), daemon=True).start()

# Load the model when the server starts
if __name__ == "__main__":
    load_model()  # Pre-load the model
    demo.launch(server_name="0.0.0.0")