app.py

# app.py
# =============
# This is a complete app.py file for a text generation app using the Qwen/Qwen2.5-Coder-0.5B-Instruct model.
# The app uses the Gradio library to create a web interface for interacting with the model.

# Imports
# =======
import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer

# Constants
# =========
MODEL_NAME = "Qwen/Qwen2.5-Coder-0.5B-Instruct"
SYSTEM_MESSAGE = "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."

# Load Model and Tokenizer
# ========================
def load_model_and_tokenizer():
    """
    Load the model and tokenizer from Hugging Face.
    """
    tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
    model = AutoModelForCausalLM.from_pretrained(
        MODEL_NAME,
        torch_dtype="auto",
        device_map="cpu"  # Ensure the model runs on the CPU
    )
    return model, tokenizer

# Ensure the model and tokenizer are loaded
model, tokenizer = load_model_and_tokenizer()

# Generate Response
# =================
def generate_response(prompt, chat_history, max_new_tokens, temperature):
    """
    Generate a response from the model based on the user prompt and chat history.
    """
    messages = [{"role": "system", "content": SYSTEM_MESSAGE}] + chat_history + [{"role": "user", "content": prompt}]
    text = tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True
    )
    model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

    generated_ids = model.generate(
        **model_inputs,
        max_new_tokens=max_new_tokens,
        do_sample=True,
        top_k=50,
        top_p=0.95,
        temperature=temperature,
        output_scores=True,
        return_dict_in_generate=True,
        return_legacy_cache=True  # Ensure legacy format is returned
    )

    response = ""
    for token_id in generated_ids.sequences[0][len(model_inputs.input_ids[0]):]:
        response += tokenizer.decode([token_id], skip_special_tokens=True)
        yield response

# Clear Chat History
# ==================
def clear_chat():
    """
    Clear the chat history.
    """
    return []

# Gradio Interface
# =================
def gradio_interface():
    """
    Create and launch the Gradio interface.
    """
    with gr.Blocks() as demo:
        with gr.Row():
            with gr.Column(scale=3):
                chatbot = gr.Chatbot(label="Chat with Qwen/Qwen2.5-Coder-0.5B-Instruct", type="messages")
                msg = gr.Textbox(label="User Input")
                with gr.Row():
                    submit = gr.Button("Submit")
                    clear = gr.Button("Clear Chat")
            with gr.Column(scale=1):
                with gr.Group():
                    gr.Markdown("### Settings")
                    max_new_tokens = gr.Slider(50, 1024, value=512, step=1, label="Max New Tokens")
                    temperature = gr.Slider(0.1, 1.0, value=0.7, step=0.05, label="Temperature")

        def respond(message, chat_history, max_new_tokens, temperature):
            chat_history.append({"role": "user", "content": message})
            response = ""
            for chunk in generate_response(message, chat_history, max_new_tokens, temperature):
                response = chunk
                yield chat_history + [{"role": "assistant", "content": response}]
            chat_history.append({"role": "assistant", "content": response})

        submit.click(respond, [msg, chatbot, max_new_tokens, temperature], [chatbot])
        clear.click(clear_chat, None, [chatbot])

    demo.launch()

# Main
# ====
if __name__ == "__main__":
    gradio_interface()

# Dependencies
# =============
# The following dependencies are required to run this app:
# - transformers
# - gradio
# - torch
# - accelerate
#
# You can install these dependencies using pip:
# pip install transformers gradio torch accelerate