chat_llama.py

import torch
from peft import PeftModel    
from transformers import AutoModelForCausalLM, AutoTokenizer, LlamaTokenizer, StoppingCriteria, StoppingCriteriaList, TextIteratorStreamer

model_name = "../llama/llama_weight/Llama-2-7b-hf"
adapters_name = '../ctranslate2/checkpoint/base'

print(f"Starting to load the model {model_name} into memory")

m = AutoModelForCausalLM.from_pretrained(
    model_name,
    # load_in_8bit=True,
    torch_dtype=torch.float16,
    device_map="auto"
)
print("finishend load model")
m = PeftModel.from_pretrained(m, adapters_name)
m = m.merge_and_unload()
print("finished merge model")
tok = LlamaTokenizer.from_pretrained(model_name)
tok.model_max_length=8192
# tok.pad_token_id = 0

stop_token_ids = [0]

print(f"Successfully loaded the model {model_name} into memory")


import datetime
import os
from threading import Event, Thread
from uuid import uuid4

import gradio as gr
import requests

max_new_tokens = 1536
start_message = """A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions."""

ORCA_PROMPT_DICT={"prompt_no_input":(
    "### System:\n"
    "You are an AI assistant that follows instruction extremely well. Help as much as you can."
    "\n\n### User:\n"
),
"prompt_input":(
    "### System:\n"
    "You are an AI assistant that follows instruction extremely well. Help as much as you can.\n\n"
    "### User:\n"
    "{instruction}"
    "\n\n### Input:\n"
    "{input}"
    "\n\n### Response:"
)}

ORCA_PROMPT_DICT={"prompt_no_input":(
        "### System:\n"
        "You are an AI assistant that follows instruction extremely well. Help as much as you can.")
    }

PROMPT_DICT = {
        "prompt_input": (
            "Below is an instruction that describes a task, paired with an input that provides further context. "
            "Write a response that appropriately completes the request.\n\n"
            "### Instruction:\n{instruction}\n\n### Input:\n{input}\n\n### Response:"
        ),
        "prompt_no_input": (
            "Below is an instruction that describes a task. "
            "Write a response that appropriately completes the request.\n\n"
            "{instruction}\n\n### Response:"
        ),
        }


llama2_prompt ={ "prompt_no_input":"""[INST] <<SYS>>
You are a helpful, respectful and honest assistant.Help as much as you can.
<</SYS>>

{instruction} [/INST] """}

class StopOnTokens(StoppingCriteria):
    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
        for stop_id in stop_token_ids:
            if input_ids[0][-1] == stop_id:
                return True
        return False


def convert_history_to_text(history):
    if len(history) > 10:
        print("*"*30)
        print("回话超过10轮，重新启动新的会话")
        history = history[10:]
    # text = llama2_prompt['prompt_no_input'] + "".join(
    #     [
    #         "".join(
    #             [
    #                 # f"### Human: {item[0]}\n",
    #                 # f"### Assistant: {item[1]}\n",
    #                 # f"USER: {item[0]}",
    #                 #ASSISTANT: {item[1]}
    #                 # f"\n\n### User:\n{item[0]}",
    #                 # f"\n\n### Response:{item[1]}"
    #                 # f"### Instruction:\n{item[0]}\n\n",
    #                 # f"### Response:{item[0]}"
                    
                    
                    
    #             ]
    #         )
    #         for item in history[:-1]
    #     ]
    # )
    # text += "".join(
    #     [
    #         "".join(
    #             [
    #                 # f"### Human: {history[-1][0]}\n",
    #                 # f"### Assistant: {history[-1][1]}\n",
    #                 # f"USER: {history[-1][0]}",
    #                 #"ASSISTANT: {history[-1][1]}"
    #                  # f"\n\n### User:\n{history[-1][0]}",
    #                 # f"\n\n### Response:{history[-1][1]}"
    #                 f"### Instruction:\n{history[-1][0]}\n\n",
    #                 f"### Response:{history[-1][1]}"
    #             ]
    #         )
    #     ]
    # )
    start_msg = llama2_prompt['prompt_no_input'].format_map({"instruction":history[0][0]})
    if len(history) > 1:
        start_msg = start_msg + history[0][1] + "</s>"
    for dialogue_his in history[1:-1]:
        start_msg += f"<s>[INST] {dialogue_his[0]}[/INST]"
        start_msg += f"{dialogue_his[1]}</s>"
    if len(history) > 1:
        start_msg += f"<s> [INST] {history[-1][0]} [/INST]"
    print(f"input msg:{start_msg}")
    return start_msg


def log_conversation(conversation_id, history, messages, generate_kwargs):
    logging_url = os.getenv("LOGGING_URL", None)
    if logging_url is None:
        return

    timestamp = datetime.datetime.now().strftime("%Y-%m-%dT%H:%M:%S")

    data = {
        "conversation_id": conversation_id,
        "timestamp": timestamp,
        "history": history,
        "messages": messages,
        "generate_kwargs": generate_kwargs,
    }

    try:
        print(f"data:{data}")
        requests.post(logging_url, json=data)
    except requests.exceptions.RequestException as e:
        print(f"Error logging conversation: {e}")


def user(message, history):
    # Append the user's message to the conversation history
    return "", history + [[message, ""]]


def bot(history, temperature, top_p, top_k, repetition_penalty, conversation_id):
    print(f"history: {history}")
    # Initialize a StopOnTokens object
    stop = StopOnTokens()

    # Construct the input message string for the model by concatenating the current system message and conversation history
    messages = convert_history_to_text(history)

    # Tokenize the messages string
    input_ids = tok(messages, return_tensors="pt").input_ids
    input_ids = input_ids.to(m.device)
    streamer = TextIteratorStreamer(tok, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        input_ids=input_ids,
        max_new_tokens=max_new_tokens,
        temperature=temperature,
        do_sample=temperature > 0.0,
        top_p=top_p,
        top_k=top_k,
        num_beams=1,
        repetition_penalty=repetition_penalty,
        streamer=streamer,
        stopping_criteria=StoppingCriteriaList([stop]),
    )

    stream_complete = Event()

    def generate_and_signal_complete():
        m.generate(**generate_kwargs)
        stream_complete.set()

    def log_after_stream_complete():
        stream_complete.wait()
        log_conversation(
            conversation_id,
            history,
            messages,
            {
                "top_k": top_k,
                "top_p": top_p,
                "temperature": temperature,
                "repetition_penalty": repetition_penalty,
            },
        )

    t1 = Thread(target=generate_and_signal_complete)
    t1.start()

    t2 = Thread(target=log_after_stream_complete)
    t2.start()

    # Initialize an empty string to store the generated text
    partial_text = ""
    for new_text in streamer:
        partial_text += new_text
        history[-1][1] = partial_text
        yield history


def get_uuid():
    return str(uuid4())


with gr.Blocks(
    theme=gr.themes.Soft(),
    css=".disclaimer {font-variant-caps: all-small-caps;}",
) as demo:
    conversation_id = gr.State(get_uuid)
    gr.Markdown(
        """得物客服智能机器人
"""
    )
    chatbot = gr.Chatbot().style(height=500)
    with gr.Row():
        with gr.Column():
            msg = gr.Textbox(
                label="Chat Message Box",
                placeholder="聊天输入框",
                show_label=False,
            ).style(container=False)
        with gr.Column():
            with gr.Row():
                submit = gr.Button("Submit")
                stop = gr.Button("Stop")
                clear = gr.Button("Clear")
    with gr.Row():
        with gr.Accordion("Advanced Options:", open=False):
            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        temperature = gr.Slider(
                            label="Temperature",
                            value=0.8,
                            minimum=0.0,
                            maximum=1.0,
                            step=0.1,
                            interactive=True,
                            info="Higher values produce more diverse outputs",
                        )
                with gr.Column():
                    with gr.Row():
                        top_p = gr.Slider(
                            label="Top-p (nucleus sampling)",
                            value=0.83,
                            minimum=0.0,
                            maximum=1,
                            step=0.01,
                            interactive=True,
                            info=(
                                "Sample from the smallest possible set of tokens whose cumulative probability "
                                "exceeds top_p. Set to 1 to disable and sample from all tokens."
                            ),
                        )
                with gr.Column():
                    with gr.Row():
                        top_k = gr.Slider(
                            label="Top-k",
                            value=4,
                            minimum=0.0,
                            maximum=200,
                            step=1,
                            interactive=True,
                            info="Sample from a shortlist of top-k tokens — 0 to disable and sample from all tokens.",
                        )
                with gr.Column():
                    with gr.Row():
                        repetition_penalty = gr.Slider(
                            label="Repetition Penalty",
                            value=1.3,
                            minimum=1.0,
                            maximum=2.0,
                            step=0.1,
                            interactive=True,
                            info="Penalize repetition — 1.0 to disable.",
                        )
                # with gr.Column():
                #     with gr.Row():
                #         repetition_penalty = gr.Slider(
                #             label="beam_size",
                #             value=3,
                #             minimum=1,
                #             maximum=10,
                #             step=1,
                #             interactive=True,
                #             info="Penalize repetition — 1.0 to disable.",
                #         )
    with gr.Row():
        gr.Markdown(
            "免责声明：该模型可能会产生与事实不符的输出，不应依赖该模型来产生与事实相符的信息。模型在各种公共数据集以及得物一些商品信息进行训练。尽管做了大量的数据清洗，但是模型的输出结果还可能存在一些问题",
            elem_classes=["disclaimer"],
        )
    with gr.Row():
        gr.Markdown(
            "算法二组",
            elem_classes=["disclaimer"],
        )

    submit_event = msg.submit(
        fn=user,
        inputs=[msg, chatbot],
        outputs=[msg, chatbot],
        queue=False,
    ).then(
        fn=bot,
        inputs=[
            chatbot,
            temperature,
            top_p,
            top_k,
            repetition_penalty,
            conversation_id,
        ],
        outputs=chatbot,
        queue=True,
    )
    submit_click_event = submit.click(
        fn=user,
        inputs=[msg, chatbot],
        outputs=[msg, chatbot],
        queue=False,
    ).then(
        fn=bot,
        inputs=[
            chatbot,
            temperature,
            top_p,
            top_k,
            repetition_penalty,
            conversation_id,
        ],
        outputs=chatbot,
        queue=True,
    )
    stop.click(
        fn=None,
        inputs=None,
        outputs=None,
        cancels=[submit_event, submit_click_event],
        queue=False,
    )
    clear.click(lambda: None, None, chatbot, queue=False)


demo.queue(max_size=128, concurrency_count=2)

demo.launch(share=True)