Add download model to a separate file

Dockerfile

@@ -12,7 +12,7 @@ WORKDIR /app
 COPY . .
 
 # Install dependencies
-RUN pip install --no-cache-dir -r requirements.txt 
+RUN pip install --no-cache-dir -r requirements.txt
 
 # Set Hugging Face cache directory to persist model downloads
 ENV HF_HOME="/home/user/.cache/huggingface"
@@ -22,22 +22,14 @@ ENV SENTENCE_TRANSFORMERS_HOME="/home/user/.cache/huggingface/sentence-transform
 RUN mkdir -p $SENTENCE_TRANSFORMERS_HOME && \
     chown -R user:user /home/user/.cache/huggingface
 
-# Download and persist the model properly (USE snapshot_download)
-RUN python -c "import os, shutil; \
-    from huggingface_hub import snapshot_download; \
-    model_path = snapshot_download(repo_id='sentence-transformers/all-MiniLM-L6-v2', cache_dir='/app/model_cache'); \
-    snapshot_dir = os.path.join(model_path, 'snapshots', os.listdir(os.path.join(model_path, 'snapshots'))[0]); \
-    for filename in os.listdir(snapshot_dir): \
-        shutil.move(os.path.join(snapshot_dir, filename), '/app/model_cache/')"
-
-# Ensure the model files are available at runtime (list out)
-RUN ls -l /app/model_cache && cat /app/model_cache/config.json
+# Run the model download script
+RUN python /app/download_model.py
 
 # Ensure ownership and permissions remain intact
-RUN ls -l /app/model_cache
+RUN chown -R user:user /app/model_cache
 
-# Expose application port
+# Expose port
 EXPOSE 7860
 
-# Run server
+# Run the application
 CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -69,11 +69,11 @@ os.environ["SENTENCE_TRANSFORMERS_HOME"] = hf_cache_dir
 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...")
-# 4a) First, try loading from our copied cache
-if os.path.exists(os.path.join(model_cache_dir, "config.json")):
+# a) First, try loading from our copied cache
+if os.path.exists(model_cache_dir) and os.listdir(model_cache_dir):  # Check if model folder exists and is not empty
     print(f"✅ Found cached model at {model_cache_dir}")
     model_loc = model_cache_dir
-# 4b) Else, try loading backup from snapshot_download
+# b) Else, try loading backup from snapshot_download
 else:
     print(f"❌ Model not found in {model_cache_dir}. This should not happen!")
     print("⚠️ Retrying with snapshot_download...")
@@ -82,7 +82,7 @@ else:
         cache_dir=hf_cache_dir,
         local_files_only=True # Change to `False` for fallback to online download
     )
-# 5. Load the model to application
+# 4. Load the model to application
 from sentence_transformers import SentenceTransformer
 print("📥 **Loading Embedding Model...**")
 # st.write("📥 **Loading Embedding Model...**")

download_model.py

@@ -0,0 +1,21 @@
+import os
+import shutil
+from huggingface_hub import snapshot_download
+
+# Define the target cache directory
+MODEL_CACHE_DIR = "/app/model_cache"
+
+# Download model
+print("⏳ Downloading the SentenceTransformer model...")
+model_path = snapshot_download(repo_id="sentence-transformers/all-MiniLM-L6-v2", cache_dir=MODEL_CACHE_DIR)
+
+# Find the snapshot folder
+snapshots_dir = os.path.join(model_path, "snapshots")
+if os.path.exists(snapshots_dir):
+    snapshot_subdirs = os.listdir(snapshots_dir)
+    if snapshot_subdirs:
+        snapshot_dir = os.path.join(snapshots_dir, snapshot_subdirs[0])
+        # Move all files to the main model cache directory
+        for filename in os.listdir(snapshot_dir):
+            shutil.move(os.path.join(snapshot_dir, filename), MODEL_CACHE_DIR)
+print(f"✅ Model downloaded and stored in {MODEL_CACHE_DIR}")