deploy

services/embedding.py

@@ -5,18 +5,16 @@ from typing import List, Optional
 import torch
 import os
 
-# Set multiple environment variables to redirect all caching
-os.environ["TRANSFORMERS_CACHE"] = "/app/cache"
-os.environ["HF_HOME"] = "/app/cache"
-os.environ["XDG_CACHE_HOME"] = "/app/cache"  # Additional variable for broader compatibility
+from huggingface_hub import HfFileSystem
+import transformers.utils.hub
 
-# Ensure cache directory exists (redundant with Dockerfile but added for safety)
-cache_dir = "/app/cache"
-if not os.path.exists(cache_dir):
-    os.makedirs(cache_dir, exist_ok=True)
+# Set cache directory globally for huggingface_hub
+transformers.utils.hub.HUGGINGFACE_HUB_CACHE = "/app/cache/hub"
 
-# Load model with custom cache directory
+# Load model
 model = SentenceTransformer("clip-ViT-B-32")
+# Load model with custom cache directory
+# model = SentenceTransformer("clip-ViT-B-32")
 
 def get_text_embedding(text: str) -> Optional[List[float]]:
     try: