3 layers for model loading

Dockerfile

@@ -18,10 +18,17 @@ RUN pip install --no-cache-dir -r requirements.txt
 ENV HF_HOME="/home/user/.cache/huggingface"
 ENV SENTENCE_TRANSFORMERS_HOME="/home/user/.cache/huggingface/sentence-transformers"
 
+# Create cache directories and ensure permissions
+RUN mkdir -p $SENTENCE_TRANSFORMERS_HOME && \
+    chown -R user:user /home/user/.cache/huggingface
+
 # Download and persist SentenceTransformer model during build stage
 RUN python -c "from sentence_transformers import SentenceTransformer; \
     SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2', cache_folder='/home/user/.cache/huggingface/sentence-transformers')"
 
+# Copy model to ensure availability at runtime
+RUN cp -r /home/user/.cache/huggingface/sentence-transformers /app/model_cache
+
 # Ensure ownership and permissions remain intact
 RUN chown -R user:user /home/user/.cache/huggingface
 

app.py

@@ -62,25 +62,32 @@ os.environ["TOKENIZERS_PARALLELISM"] = "false"
 hf_cache_dir = "/home/user/.cache/huggingface"
 os.environ["HF_HOME"] = hf_cache_dir
 os.environ["SENTENCE_TRANSFORMERS_HOME"] = hf_cache_dir
+# Model storage location
+model_cache_dir = "/app/model_cache"
 # 3. Download (or load from cache) the SentenceTransformer model 
 from huggingface_hub import snapshot_download
 print("⏳ Checking or downloading the all-MiniLM-L6-v2 model from huggingface_hub...")
 # st.write("⏳ Checking or downloading the all-MiniLM-L6-v2 model from huggingface_hub...")
-try:
-    model_loc = snapshot_download(
-        repo_id="sentence-transformers/all-MiniLM-L6-v2",
-        cache_dir=hf_cache_dir, # Set directly also fine `os.environ["HF_HOME"]`
-        local_files_only=True   # 🚨 Avoids re-downloading / fetch from internet (set False for local dev)
-    )
-except Exception as e:
-    print(f"❌ Error loading model from cache: {e}")
-    print("⚠️ Retrying with online download enabled...")
-    model_loc = snapshot_download(
-        repo_id="sentence-transformers/all-MiniLM-L6-v2",
-        cache_dir=hf_cache_dir,
-        local_files_only=False   # ⬇️ Fallback
-    )
-print(f"✅ Model directory: {model_loc}")
+if os.path.exists(model_cache_dir):
+    print(f"✅ Found cached model at {model_cache_dir}")
+    model_loc = model_cache_dir
+else: 
+    try:
+        model_loc = snapshot_download(
+            repo_id="sentence-transformers/all-MiniLM-L6-v2",
+            cache_dir=hf_cache_dir, # Set directly also fine `os.environ["HF_HOME"]`
+            local_files_only=True   # 🚨 Avoids re-downloading / fetch from internet (set False for local dev)
+        )
+        print(f"✅ Model loaded from local cache: {model_loc}")
+    except Exception as e:
+        print(f"❌ Error loading model from cache: {e}")
+        print("⚠️ Retrying with online download enabled...")
+        model_loc = snapshot_download(
+            repo_id="sentence-transformers/all-MiniLM-L6-v2",
+            cache_dir=hf_cache_dir,
+            local_files_only=False   # ⬇️ Fallback
+        )
+        print(f"✅ Model directory reloaded: {model_loc}")
 # st.write(f"✅ Model directory: {model_loc}")
 
 from sentence_transformers import SentenceTransformer