Fixed model loading

app.py

@@ -4,12 +4,17 @@ from pydantic import BaseModel
 import torch
 import joblib
 import os
+import nltk
 from transformers import AutoTokenizer
 from model_inference import load_model, predict_from_input, download_from_hf
 
-# ✅ FIX: Set Hugging Face cache to a writable directory
+# ✅ Redirect HuggingFace & NLTK cache
 os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf_cache"
 os.environ["HF_HOME"] = "/tmp/hf_cache"
+nltk_data_dir = "/tmp/nltk_data"
+os.makedirs(nltk_data_dir, exist_ok=True)
+nltk.data.path.append(nltk_data_dir)
+os.environ["NLTK_DATA"] = nltk_data_dir
 
 app = FastAPI(title="Question Difficulty/Discrimination Predictor")