app.py

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

# Load the Hugging Face API token from environment variables
hf_token = os.getenv("HF_TOKEN")

# Model name
model_name = "meta-llama/Llama-2-7b-chat-hf"

# Load the model and tokenizer with the token
tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=hf_token)
model = AutoModelForCausalLM.from_pretrained(model_name, use_auth_token=hf_token)

# Define the chat function
def chat_with_llama2(input_text):
    inputs = tokenizer(input_text, return_tensors="pt")
    outputs = model.generate(inputs["input_ids"], max_length=512, do_sample=True, top_p=0.95, top_k=60)
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return response

# Create the Gradio interface
interface = gr.Interface(
    fn=chat_with_llama2,
    inputs="text",
    outputs="text",
    title="LLaMa 2 Chat HF",
    description="Chat with LLaMa 2 model using Hugging Face Transformers and Gradio.",
    examples=[
        ["Hello, LLaMa 2! How are you today?"],
        ["Can you tell me a joke?"],
        ["What is the capital of France?"]
    ]
)

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