app.py

import os
from huggingface_hub import login
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
import gradio as gr

# Retrieve the Hugging Face token from the environment variable
hf_token = os.getenv('HF_TOKEN')

# Authenticate using the token
if hf_token:
    login(token=hf_token)
else:
    print("Error: Hugging Face token not found.")
    exit(1)

# Load the model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Meta-Llama-3.1-70B")
model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Meta-Llama-3.1-70B",
    rope_scaling={'type': 'llama3', 'factor': 8.0}  # Adjust the type and factor as needed
)

# Create a text generation pipeline
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

# Load sentiment analysis pipeline
sentiment_analyzer = pipeline("sentiment-analysis")

# Initialize conversation context
context = []

def predict(context, input_text):
    """Generate response based on context and input."""
    context.append(input_text)
    inputs = tokenizer(" ".join(context), return_tensors="pt")
    outputs = model.generate(inputs.input_ids, max_length=200, pad_token_id=tokenizer.eos_token_id)
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    context.append(response)
    return response

def predict_with_emotion(context, input_text):
    """Generate response with emotion detection."""
    sentiment = sentiment_analyzer(input_text)[0]['label']
    response = predict(context, input_text)
    if sentiment == 'NEGATIVE':
        response = "I'm sorry to hear that. " + response
    elif sentiment == 'POSITIVE':
        response = "That's great! " + response
    return response

def chatbot(input_text):
    """Gradio chatbot function."""
    global context
    response = predict_with_emotion(context, input_text)
    return response

# Create Gradio interface
iface = gr.Interface(fn=chatbot, inputs="text", outputs="text", title="Contextual Emotion-Aware LLaMA-70B Chatbot")
iface.launch()