app.py

import streamlit as st
from transformers import T5ForConditionalGeneration, T5Tokenizer
import torch

# Load pre-trained model and tokenizer from the "KhantKyaw/T5-small_new_chatbot"
model_name = "KhantKyaw/T5-small_new_chatbot"  # Use the fine-tuned model
model = T5ForConditionalGeneration.from_pretrained(model_name)
tokenizer = T5Tokenizer.from_pretrained(model_name)

# Set device to GPU if available for faster inference, otherwise fallback to CPU
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)

# Streamlit Interface
st.title("Mental Health Chatbot with T5")

def generate_response(input_text):
    # Add conversational context to input
    input_text = f"You are a helpful assistant. {input_text}"

    # Tokenize input text
    input_ids = tokenizer.encode(input_text, return_tensors="pt").to(device)
    
    # Generate a response from the model with advanced generation settings
    outputs = model.generate(input_ids,
                             max_length=100,  # max length of the output sequence
                             num_beams=5,  # Beam search for better results
                             top_p=0.95,  # Top-p sampling for more variety
                             temperature=0.7,  # Temperature controls randomness
                             no_repeat_ngram_size=2,  # Prevent repetition of n-grams
                             pad_token_id=tokenizer.eos_token_id)  # Padding token to avoid padding tokens being part of the output
    
    # Decode the model's output to a readable string
    bot_output = tokenizer.decode(outputs[0], skip_special_tokens=True)

    return bot_output

# Create input box for user to type a message
user_input = st.text_input("You: ", "")

if user_input:
    # Generate and display the bot's response
    response = generate_response(user_input)
    st.write(f"Bot: {response}")