llm chat

llmchat.py

# based on StableLM chat
# https://huggingface.co/spaces/stabilityai/stablelm-tuned-alpha-chat
import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline, StoppingCriteria, StoppingCriteriaList, TextIteratorStreamer
import time
import numpy as np
from torch.nn import functional as F
import os
from threading import Thread

print(f"Starting to load the model to memory")

model_name = "line-corporation/japanese-large-lm-1.7b-instruction-sft"
#model_name = "line-corporation/japanese-large-lm-3.6b-instruction-sft"


m = AutoModelForCausalLM.from_pretrained(model_name)
tok = AutoTokenizer.from_pretrained(model_name, use_fast = False)
generator = pipeline('text-generation', model=m, tokenizer=tok)
print(f"Sucessfully loaded the model to the memory")

start_message = ""

class StopOnTokens(StoppingCriteria):
    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
        stop_ids = [50278, 50279, 50277, 1, 0]
        for stop_id in stop_ids:
            if input_ids[0][-1] == stop_id:
                return True
        return False


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


def chat(curr_system_message, history):
    # Initialize a StopOnTokens object
    stop = StopOnTokens()

    # Construct the input message string for the model by concatenating the current system message and conversation history
    curr_system_message = ""
    messages = curr_system_message \
        .join(["\n".join(["ユーザー: "+item[0], "システム: "+item[1]])
            for item in history])
    # Tokenize the messages string
    model_inputs = tok.encode(messages, return_tensors="pt", add_special_tokens=False)
    streamer = TextIteratorStreamer(
        tok, timeout=10., skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
#        model_inputs,
        streamer=streamer,
        max_new_tokens=1024,
        do_sample=True,
        top_p=0.95,
        top_k=1000,
        temperature=0.7,
        num_beams=1,
        pad_token_id=tok.pad_token_id,
        bos_token_id=tok.bos_token_id,
        eos_token_id=tok.eos_token_id,
        no_repeat_ngram_size=2,
        stopping_criteria=StoppingCriteriaList([stop])
    )
    t = Thread(target=m.generate, args=[model_inputs], kwargs=generate_kwargs)
    t.start()

    # print(history)
    # Initialize an empty string to store the generated text
    partial_text = ""
    for new_text in streamer:
        # print(new_text)
        partial_text += new_text
        history[-1][1] = partial_text
        # Yield an empty string to cleanup the message textbox and the updated conversation history
        yield history
    return partial_text


with gr.Blocks() as demo:
    # history = gr.State([])
    gr.Markdown(f"## {model_name} Chat")
    chatbot = gr.Chatbot().style(height=500)
    with gr.Row():
        with gr.Column():
            msg = gr.Textbox(label="Chat Message Box", placeholder="Chat Message Box",
                             show_label=False).style(container=False)
        with gr.Column():
            with gr.Row():
                submit = gr.Button("Submit")
                stop = gr.Button("Stop")
                clear = gr.Button("Clear")
    system_msg = gr.Textbox(
        start_message, label="System Message", interactive=False, visible=False)

    submit_event = msg.submit(fn=user, inputs=[msg, chatbot], outputs=[msg, chatbot], queue=False).then(
        fn=chat, inputs=[system_msg, chatbot], outputs=[chatbot], queue=True)
    submit_click_event = submit.click(fn=user, inputs=[msg, chatbot], outputs=[msg, chatbot], queue=False).then(
        fn=chat, inputs=[system_msg, chatbot], 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=32, concurrency_count=2)
demo.launch()