app.py

import os
import torch
from flask import Flask, jsonify, request
from flask_cors import CORS
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, GenerationConfig
import re

# Set the HF_HOME environment variable to a writable directory
os.environ["HF_HOME"] = "/workspace/huggingface_cache"

app = Flask(__name__)

# Enable CORS for specific origins
CORS(app, resources={r"/send_message": {"origins": ["http://localhost:3000", "https://main.dbn2ikif9ou3g.amplifyapp.com"]}})

# Global variables for model and tokenizer
model = None
tokenizer = None

def get_model_and_tokenizer(model_id: str):
    global model, tokenizer
    if model is None or tokenizer is None:
        try:
            print(f"Loading tokenizer for model_id: {model_id}")
            tokenizer = AutoTokenizer.from_pretrained(model_id)
            tokenizer.pad_token = tokenizer.eos_token

            print(f"Loading model for model_id: {model_id}")
            
            bnb_config = BitsAndBytesConfig(
                load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.float16, bnb_4bit_use_double_quant=True
            )
            
            model = AutoModelForCausalLM.from_pretrained(
                model_id, quantization_config=bnb_config, device_map="auto"
            )
            
            model.config.use_cache = False
            model.config.pretraining_tp = 1
            model.config.pad_token_id = tokenizer.eos_token_id  # Fix padding issue

        except Exception as e:
            print(f"Error loading model: {e}")
            raise e

def generate_response(user_input, model_id):
    # Ensure model and tokenizer are loaded
    get_model_and_tokenizer(model_id)

    prompt = user_input
    device = "cuda" if torch.cuda.is_available() else "cpu"
    
    generation_config = GenerationConfig(
        penalty_alpha=0.6,
        do_sample=True,
        top_p=0.2,
        top_k=50,
        temperature=0.3,
        repetition_penalty=1.2,
        max_new_tokens=60,
        pad_token_id=tokenizer.eos_token_id
    )

    inputs = tokenizer(prompt, return_tensors="pt").to(device)
    model.to(device)

    outputs = model.generate(**inputs, generation_config=generation_config)
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)

    # Clean up response
    cleaned_response = re.sub(r"(User:|Assistant:)", "", response).strip()
    return cleaned_response.split("\n")[0]  # Keep only the first line of response

@app.route("/", methods=["GET"])
def handle_get_request():
    message = request.args.get("message", "No message provided.")
    return jsonify({"message": message, "status": "GET request successful!"})

@app.route("/send_message", methods=["POST"])
def handle_post_request():
    data = request.get_json()
    if data is None:
        return jsonify({"error": "No JSON data provided"}), 400

    message = data.get("inputs", "No message provided.")
    model_id = data.get("model_id", "YALCINKAYA/FinetunedByYalcin")

    try:
        print(f"Processing request")
        model_response = generate_response(message, model_id)
        return jsonify({
            "received_message": model_response,
            "model_id": model_id,
            "status": "POST request successful!"
        })
    except Exception as e:
        error_message = str(e) if app.debug else "An error occurred while processing your request."
        print(f"Error handling POST request: {e}")
        return jsonify({"error": error_message}), 500

if __name__ == '__main__':
    app.run(host='0.0.0.0', port=7860)