app.py

from threading import Thread
import gradio as gr
import transformers
import torch
from transformers import pipeline, AutoTokenizer, TextIteratorStreamer


def chat_history(history) -> str:
    messages = [
        {
            "role": ("user" if i % 2 == 0 else "assistant"),
            "content": dialog[i % 2]
        }
        for i, dialog in enumerate(history) for _ in (0, 1) if dialog[i % 2]
    ]

    return pipeline.tokenizer.apply_chat_template(
        messages, toknizer=False, add_generation_prompt=True
    )


def model_loading_pipeline():
    model_id = "vilm/vinallama-7b"
    tokenizer = AutoTokenizer.from_pretrained(model_id)
    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, Timeout=5)
    pipeline = transformers.pipeline(
        "text-generation",
        model=model_id,
        model_kwargs={
            "torch_dtype": torch.float16,
            "load_in_8bits": True,
        },
        streamer=streamer,
    )
    return pipeline, streamer


def launch_app(pipeline, streamer):
    with gr.Blocks() as demo:
        chat = gr.Chatbot()
        msg = gr.Textbox()
        clear = gr.Button("Clear")

        def user(user_message, history):
            return "", history + [[user_message, None]]

        def bot(history):
            prompt = chat_history(history)
            history[-1][1] = ""
            kwargs = {
                "text_inputs": prompt,
                "max_new_tokens": 2048,
                "do_sample": True,
                "temperature": 0.7,
                "top_k": 50,
                "top_p": 0.95,
            }
            thread = Thread(target=pipeline, kwargs=kwargs)
            thread.start()

            for token in streamer:
                history[-1][1] += token
                yield history

        msg.submit(user, [msg, chat], [msg, chat], queue=False).then(bot, chat, chat)
        clear.click(lambda: None, None, chat, queue=False)

    demo.queue()
    demo.launch(share=True, debug=True)


if __name__ == "__main__":
    pipe, streamer = model_loading_pipeline()
    launch_app(pipe, streamer)