app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
import accelerate

model_path = "01-ai/Yi-6B-200K"
model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained(model_path)

def respond_to_input(user_input):
    prompt = f"""
    You are sahaAI, an empathetic chatbot, which consoles the User, and makes his/her mood positive, joyful and enjoyable.
    <SahaAI>: Hello! I'm here to help you with any psychological concerns or questions you may have. Feel free to share your thoughts and feelings with me, and I'll do my best to provide support and guidance.
    <User>: {user_input}
    <SahaAI>:
    """
    inputs = tokenizer(prompt, return_tensors="pt")
    max_length = 256
    outputs = model.generate(
        inputs.input_ids,
        max_length=max_length,
        eos_token_id=tokenizer.eos_token_id,
        do_sample=True,
        repetition_penalty=1.3,
        no_repeat_ngram_size=5,
        temperature=0.5,
        top_k=40,
        top_p=0.8,
    )
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

interface = gr.Interface(
    fn=respond_to_input,
    inputs=gr.Textbox(lines=5, placeholder="Type your message here..."),  # Multiline text input
    outputs="text",  # Text output
    title="SahaAI",  # Title of the web interface
    description="Online psychology assistant, to help your mental health.",  # Description of the web interface
    theme="dark"  # Dark theme to match your screenshot
)

interface.launch()