app.py

import os
import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer

TITLE = "Chattybot"
EXAMPLE_INPUT = "hello"
SYSTEM_PROMPT = "As a generative chatbot (you are not a GPT but your structure is 50% the same), your primary function is to provide helpful and friendly responses to user queries. Feel free to add some personality, but make sure your responses are accurate and helpful. Your owner and developer is: @Costikoooo (Discord user) other developers are unknown. Your name is Chattybot."

model_name = "HuggingFaceH4/zephyr-7b-beta"
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

def predict_local(message, chatbot=[], system_prompt=""):
    input_prompt = "\n" + system_prompt + "</s>\n\n"
    for interaction in chatbot:
        input_prompt = input_prompt + str(interaction[0]) + "</s>\n\n" + str(interaction[1]) + "\n</s>\n\n"

    input_prompt = input_prompt + str(message) + "</s>\n"

    inputs = tokenizer(input_prompt, return_tensors="pt")
    outputs = model(**inputs)
    generated_text = tokenizer.decode(outputs["logits"][0], skip_special_tokens=True)
    
    return generated_text

def test_preview_chatbot(message, history):
    response = predict_local(message, history, SYSTEM_PROMPT)
    text_start = response.rfind("") + len("")
    response = response[text_start:]
    return response

welcome_preview_message = f"""
Welcome to **{TITLE}**! Say something like: 
"{EXAMPLE_INPUT}"
"""

chatbot_preview = gr.Chatbot(layout="panel", value=[(None, welcome_preview_message)])
textbox_preview = gr.Textbox(scale=7, container=False, value=EXAMPLE_INPUT)

demo = gr.ChatInterface(test_preview_chatbot, chatbot=chatbot_preview, textbox=textbox_preview)

demo.launch()