app.py

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

# Load model and tokenizer from Hugging Face Model Hub
model_name = "meta-llama/Meta-Llama-3.1-70B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Define system instruction
system_instruction = "You are a helpful assistant. Provide detailed and accurate responses to the user's queries."

# Define the chat function
def chat_function(prompt):
    # Create the full input prompt including the system instruction
    full_prompt = f"{system_instruction}\nUser: {prompt}\nAssistant:"
    
    # Tokenize the full prompt
    inputs = tokenizer(full_prompt, return_tensors="pt")
    
    # Generate model response
    with torch.no_grad():
        outputs = model.generate(**inputs, max_length=150, num_return_sequences=1)
    
    # Decode and return response
    response = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()
    
    # Extract only the assistant's response
    response = response.split("Assistant:")[-1].strip()
    return response

# Create Gradio interface
iface = gr.Interface(
    fn=chat_function,
    inputs="text",
    outputs="text",
    title="Meta-Llama Chatbot",
    description="A chatbot powered by the Meta-Llama-3.1-70B-Instruct model."
)

# Launch the interface
if __name__ == "__main__":
    iface.launch()