app.py

import gradio as gr
import spaces
import torch
from threading import Thread

from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer


TITLE = "<center>Chat with Llama3-8B-Chinese</center>"

DESCRIPTION = "<h3><center>Visit <a href='https://huggingface.co/shenzhi-wang/Llama3-8B-Chinese-Chat' target='_blank'>our model page</a> for details.</center></h3>"

TOOL_EXAMPLE = '''You have access to the following tools:
```python
def google_search(keywords: List[str]):
    """
    Search on the Internet based on the keywords.
    
    Args:
        keywords (List[str]): Keywords for the search engine.
    """
    pass
```

Write "Action:" followed by a list of actions in JSON that you want to call, e.g.
Action:
```json
[
    {
        "name": "tool name (one of [google_search])",
        "arguments": "the input to the tool"
    }
]
```
'''


tokenizer = AutoTokenizer.from_pretrained("shenzhi-wang/Llama3-8B-Chinese-Chat")
model = AutoModelForCausalLM.from_pretrained("shenzhi-wang/Llama3-8B-Chinese-Chat", device_map="auto")


@spaces.GPU
def stream_chat(message: str, history: list, system: str, temperature: float, max_new_tokens: int):
    conversation = [{"role": "system", "content": system}]
    for prompt, answer in history:
        conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])

    conversation.append({"role": "user", "content": message})

    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt").to(model.device)
    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)

    generate_kwargs = dict(
        input_ids=input_ids,
        streamer=streamer,
        max_new_tokens=max_new_tokens,
        temperature=temperature,
        do_sample=True,
    )
    if temperature == 0:
        generate_kwargs["do_sample"] = False

    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()

    output = ""
    for new_token in streamer:
        output += new_token
        yield output


with gr.Blocks(fill_height=True) as demo:
    gr.ChatInterface(
        fn=stream_chat,
        fill_height=True,
        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
        additional_inputs=[
            gr.Text(
                value="You are a helpful assistant.",
                label="System",
                render=False,
            ),
            gr.Slider(
                minimum=0,
                maximum=1,
                step=0.1,
                value=0.95,
                label="Temperature",
                render=False,
            ),
            gr.Slider(
                minimum=128,
                maximum=4096,
                step=1,
                value=512,
                label="Max new tokens",
                render=False,
            ),
        ],
        examples=[
            ["我的蓝牙耳机坏了，我该去看牙科还是耳鼻喉科？", "You are a helpful assistant."],
            ["在一道没有余数的除法算式里，被除数(不为零)加上除数和商的积，再除以被除数，所得的商是多少？", "You are a helpful assistant."],
            ["今日行军进展如何？", "扮演诸葛亮和我对话。"],
            ["羊驼的寿命有多久？", TOOL_EXAMPLE],
        ],
        cache_examples=False,
        title=TITLE,
        description=DESCRIPTION,
    )


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