model.py

from transformers import AutoTokenizer, AutoModelForCausalLM
import os
import torch


# Check if CUDA is available for faster inference
device = 'cuda' if torch.cuda.is_available() else 'cpu'

# Load the tokenizer and model once, outside of the function
huggingface_token = os.environ.get("KEY2")
tokenizer = AutoTokenizer.from_pretrained(
    "meta-llama/Llama-3.2-1B",
    use_auth_token=huggingface_token

)
model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.2-1B",
    use_auth_token=huggingface_token
).to(device)

def modelFeedback(ats_score, resume_data, job_description):
    """
    """

    try:
        # Tokenize the input
        input_ids = tokenizer.encode(input_prompt, return_tensors="pt").to(device)
        
        # Disable gradient calculation for faster inference
        with torch.no_grad():
            # Generate the output
            output = model.generate(
                input_ids,
                max_length=1500,
                temperature=0.01,
                pad_token_id=tokenizer.eos_token_id  # Ensure padding works properly
            )
        
        # Decode the output
        response_text = tokenizer.decode(output[0], skip_special_tokens=True)
        return response_text
    except Exception as e:
        print(f"Error during generation: {e}")