llama_models.py

import os
from transformers import AutoTokenizer, AutoModelForCausalLM # Ensure correct model class

HUGGINGFACE_API_KEY = os.getenv("HUGGINGFACE_API_KEY")
if HUGGINGFACE_API_KEY is None:
 raise ValueError("Hugging Face API key is not set. Please add it as a secret in your Hugging Face Space settings.")
print(f"Using Hugging Face API Key: {HUGGINGFACE_API_KEY}")

model = None
tokenizer = None

def load_model(model_name):
 global tokenizer, model
 if not tokenizer or not model:
 print("Loading model and tokenizer...")
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 tokenizer.pad_token = tokenizer.eos_token # Set pad_token to eos_token
 model = AutoModelForCausalLM.from_pretrained(model_name) # Ensure correct model class
 print("Model and tokenizer loaded successfully.")
 return tokenizer, model

async def process_text_local(model_name, text):
 print("Loading model and tokenizer...")
 tokenizer, model = load_model(model_name)
 print("Encoding text...")
 inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=512) # Set max_length to 512
 print("Text encoded successfully.")
 print("Generating output...")
 outputs = model.generate(**inputs, max_length=512)
 print("Output generated successfully.")
 result = tokenizer.decode(outputs[0], skip_special_tokens=True)
 print("Output decoded successfully.")
 return result