app.py

import os

os.system("pip install git+https://github.com/shumingma/transformers.git")

import threading
import torch
import torch._dynamo
torch._dynamo.config.suppress_errors = True

from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    TextIteratorStreamer,
)
import gradio as gr
import spaces

model_id = "microsoft/bitnet-b1.58-2B-4T"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
print(model.device)

@spaces.GPU
def respond(
    message: str,
    history: list[tuple[str, str]],
    system_message: str,
    max_tokens: int,
    temperature: float,
    top_p: float,
):
    """
    Generate a chat response using streaming with TextIteratorStreamer.

    Args:
        message: User's current message.
        history: List of (user, assistant) tuples from previous turns.
        system_message: Initial system prompt guiding the assistant.
        max_tokens: Maximum number of tokens to generate.
        temperature: Sampling temperature.
        top_p: Nucleus sampling probability.

    Yields:
        The growing response text as new tokens are generated.
    """
    messages = [{"role": "system", "content": system_message}]
    for user_msg, bot_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if bot_msg:
            messages.append({"role": "assistant", "content": bot_msg})
    messages.append({"role": "user", "content": message})

    prompt = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

    streamer = TextIteratorStreamer(
        tokenizer, skip_prompt=True, skip_special_tokens=True
    )
    generate_kwargs = dict(
        **inputs,
        streamer=streamer,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        do_sample=True,
    )
    thread = threading.Thread(target=model.generate, kwargs=generate_kwargs)
    thread.start()

    response = ""
    for new_text in streamer:
        response += new_text
        yield response

demo = gr.ChatInterface(
    fn=respond,
    title="Bitnet-b1.58-2B-4T Chatbot",
    description="This chat application is powered by Microsoft's SOTA Bitnet-b1.58-2B-4T and designed for natural and fast conversations.",
    examples=[
        [
            "Hello! How are you?",
            "You are a helpful AI assistant.",
            512,
            0.7,
            0.95,
        ],
        [
            "Can you code a snake game in Python?",
            "You are a helpful AI assistant.",
            2048,
            0.7,
            0.95,
        ],
    ],
    additional_inputs=[
        gr.Textbox(
            value="You are a helpful AI assistant.",
            label="System message"
        ),
        gr.Slider(
            minimum=1,
            maximum=8192,
            value=2048,
            step=1,
            label="Max new tokens"
        ),
        gr.Slider(
            minimum=0.1,
            maximum=4.0,
            value=0.7,
            step=0.1,
            label="Temperature"
        ),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)"
        ),
    ],
)

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