app.py

import argparse
from pathlib import Path

import chatglm_cpp
import gradio as gr

import urllib

DEFAULT_MODEL_PATH = "chatglm3-6b.bin"

urllib.request.urlretrieve(
    "https://huggingface.co/Braddy/chatglm3-6b-chitchat/resolve/main/q5_1.bin?download=true", 
    DEFAULT_MODEL_PATH
    )

parser = argparse.ArgumentParser()
parser.add_argument("-m", "--model", default=DEFAULT_MODEL_PATH, type=Path, help="model path")
parser.add_argument("--mode", default="chat", type=str, choices=["chat", "generate"], help="inference mode")
parser.add_argument("-l", "--max_new_tokens", default=64, type=int, help="max total output tokens")
parser.add_argument("-c", "--max_context_length", default=1024, type=int, help="max context length")
parser.add_argument("--top_k", default=40, type=int, help="top-k sampling")
parser.add_argument("--top_p", default=0.75, type=float, help="top-p sampling")
parser.add_argument("--temp", default=0.5, type=float, help="temperature")
parser.add_argument("--repeat_penalty", default=1.0, type=float, help="penalize repeat sequence of tokens")
parser.add_argument("-t", "--threads", default=0, type=int, help="number of threads for inference")
parser.add_argument("--plain", action="store_true", help="display in plain text without markdown support")
args = parser.parse_args()

pipeline = chatglm_cpp.Pipeline(args.model)
system_message = chatglm_cpp.ChatMessage(role="system", content="请你现在扮演一个软件工程师，名字叫做贺英旭。你需要以这个身份和朋友们对话。")


def postprocess(text):
    if args.plain:
        return f"<pre>{text}</pre>"
    return text


def predict(input, chatbot, max_new_tokens, top_p, temperature, messages):
    chatbot.append((postprocess(input), ""))
    messages.append(chatglm_cpp.ChatMessage(role="user", content=input))
    full_messages = [system_message] + messages

    generation_kwargs = dict(
        max_new_tokens=max_new_tokens,
        max_context_length=args.max_context_length,
        do_sample=temperature > 0,
        top_k=args.top_k,
        top_p=top_p,
        temperature=temperature,
        repetition_penalty=args.repeat_penalty,
        num_threads=args.threads,
        stream=True,
    )

    response = ""
    chunks = []
    for chunk in pipeline.chat(full_messages, **generation_kwargs):
        response += chunk.content
        chunks.append(chunk)
        chatbot[-1] = (chatbot[-1][0], postprocess(response))
        yield chatbot, messages
    messages.append(pipeline.merge_streaming_messages(chunks))

    yield chatbot, messages


def reset_user_input():
    return gr.update(value="")


def reset_state():
    return [], []

title = """
<div style="text-align: center">
    <h1>Chichat</h1>
    <p style="text-align: center;">Free feel to talk about anything :)</p>
</div>
"""


with gr.Blocks() as demo:
    gr.HTML(title)

    chatbot = gr.Chatbot(height=300, label="Check this out!")
    with gr.Row():
        with gr.Column(scale=4):
            user_input = gr.Textbox(show_label=False, placeholder="Input...", lines=8)
            submitBtn = gr.Button("Submit", variant="primary")
        with gr.Column(scale=1):
            max_new_tokens = gr.Slider(0, 512, value=args.max_new_tokens, step=1.0, label="Maximum output tokens", interactive=True)
            top_p = gr.Slider(0, 1, value=args.top_p, step=0.01, label="Top P", interactive=True)
            temperature = gr.Slider(0, 1, value=args.temp, step=0.01, label="Temperature", interactive=True)
            emptyBtn = gr.Button("Clear History")

    messages = gr.State([])

    submitBtn.click(
        predict,
        [user_input, chatbot, max_new_tokens, top_p, temperature, messages],
        [chatbot, messages],
        show_progress=True,
    )
    submitBtn.click(reset_user_input, [], [user_input])

    emptyBtn.click(reset_state, outputs=[chatbot, messages], show_progress=True)

demo.queue().launch(share=False, inbrowser=True)