app.py

import gradio as gr
import random
import time
from ctransformers import AutoModelForCausalLM
import datetime
import os


params = {
        "max_new_tokens":512,
        "stop":["<end>" ,"<|endoftext|>"],
        "temperature":0.7,
        "top_p":0.8,
        "stream":True,
        "batch_size": 8}


def save_log(task, to_save):
  with open("logs.txt", "a") as log_file:
      current_time = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
      log_file.write(f"[{current_time}] - {task}: {to_save}\n")
      print(to_save)
        

llm = AutoModelForCausalLM.from_pretrained("Aspik101/Llama-2-7b-chat-hf-pl-lora_GGML", model_type="llama")

with gr.Blocks() as demo:
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.Button("Clear")

    def user(user_message, history):
        return "", history + [[user_message, None]]

    def parse_history(hist):
        history_ = ""
        for q, a in hist:
            history_ += f"<user>: {q } \n"
            if a:
                history_ += f"<assistant>: {a} \n"
        return history_

    def bot(history):
        print("history: ",history)
        prompt =  f"Jesteś AI assystentem. Odpowiadaj po polsku. {parse_history(history)}. <assistant>:"
        print("prompt: ",prompt)
        stream = llm(prompt, **params)
        history[-1][1] = ""
        answer_save = ""
        for character in stream:
            history[-1][1] += character
            answer_save += character
            time.sleep(0.005)
            yield history

        print("answer_save: ",answer_save)
    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
        bot, chatbot, chatbot
    )
    clear.click(lambda: None, None, chatbot, queue=False)
    
demo.queue()
demo.launch()