app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load model and tokenizer
model_name = "meta-llama/Llama-2-7b-hf"

# Use Hugging Face authentication token
token = 'HUGGINGFACE_TOKEN'  # Replace this with your Hugging Face token

# Load model and tokenizer from Hugging Face
tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=token)
model = AutoModelForCausalLM.from_pretrained(model_name, use_auth_token=token)

# Function to generate response from the model
def generate_response(user_input, chat_history=None):
    if chat_history is None:
        chat_history = []
    
    # Format the input for the model
    input_text = user_input + ' '  # Add a space for separation between user input and the response

    # Encode the input
    inputs = tokenizer.encode(input_text, return_tensors="pt")

    # Generate a response from the model
    outputs = model.generate(inputs, max_length=150, num_return_sequences=1, no_repeat_ngram_size=2, pad_token_id=tokenizer.eos_token_id)

    # Decode the response
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    
    # Append the response to the chat history
    chat_history.append((user_input, response))
    
    # Return the response and updated chat history
    return response, chat_history

# Gradio Interface
def respond(user_input, chat_history=None):
    response, chat_history = generate_response(user_input, chat_history)
    return response, chat_history

# Set up Gradio interface
iface = gr.Interface(
    fn=respond,
    inputs=[
        gr.Textbox(label="Your Message", placeholder="Ask me anything!", lines=2),
        gr.State()
    ],
    outputs=[
        gr.Textbox(label="Response", lines=3),
        gr.State()
    ],
    title="Llama-2 Chatbot",
    description="Ask me anything, and I'll respond using Llama-2 model.",
    live=True
)

# Launch the Gradio interface
iface.launch()