app.py

import gradio as gr




# from huggingface_hub import InferenceClient
# """
# For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
# """
# client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")




from transformers import AutoModelForCausalLM, AutoTokenizer
import torch


class ChatClient:
    def __init__(self, model_path):
        """
        初始化客户端，加载模型和分词器到 GPU（如果可用）。
        """
        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        print(f"Using device: {self.device}")

        self.tokenizer = AutoTokenizer.from_pretrained(model_path)
        self.model = AutoModelForCausalLM.from_pretrained(model_path).to(self.device)
        self.model.eval()  # 设置为评估模式

    async def chat_completion(self, messages, max_tokens, stream=False, temperature=1.0, top_p=1.0):
        """
        生成对话回复。
        """
        # 将所有输入消息合并为一个字符串
        input_text = messages
        print(input_text)
        # 使用分词器处理输入文本
        inputs = self.tokenizer(input_text, return_tensors='pt').to(self.device)

        # 设置生成的参数
        gen_kwargs = {
            "max_length": inputs['input_ids'].shape[1] + max_tokens,
            "temperature": temperature,
            "top_p": top_p,
            "do_sample": True
        }

        # 使用生成器生成文本
        output_sequences = self.model.generate(**inputs, **gen_kwargs)

        # 解码生成的文本
        # result_text = self.tokenizer.decode(output_sequences[0], skip_special_tokens=True)
        # yield result_text

        # 解码生成的文本
        for sequence in output_sequences:
            result_text = self.tokenizer.decode(sequence, skip_special_tokens=True)
            await anyio.sleep(0)  # Yield control, simulating asynchronous operation
            yield result_text

# 创建客户端实例，指定模型路径
model_path = 'model/v3/'
client = ChatClient(model_path)






async def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # messages = [{"role": "system", "content": system_message}]
    #
    # for val in history:
    #     if val[0]:
    #         messages.append({"role": "user", "content": val[0]})
    #     if val[1]:
    #         messages.append({"role": "assistant", "content": val[1]})
    #
    # messages.append({"role": "user", "content": message})

    messages = system_message + message


    response = ""

    async for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        # print(message)
        # token = message
        # #token = message.choices[0].delta.content

        # response += token
        # yield response

        yield message

"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="Yahoo!ショッピングについての質問を回答してください。", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=1024, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.1, 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()