Update entrypoint.sh file

Author: bglar

pipelines/Dockerfile.terratorch

@@ -55,13 +55,14 @@ WORKDIR /terratorch
 
 ENV HF_HOME=/app/models/huggingface_cache \
     TERRATORCH_SEGMENTATION_IO_PROCESSOR_CONFIG='{"output_path": "/data/output/"}' \
-    PATH="/home/appuser/.local/bin:${PATH}"
+    PATH="/home/appuser/.local/bin:${PATH}" \
+    PYTHONUNBUFFERED=1 \
+    PYTHONDONTWRITEBYTECODE=1
 
 # Install Python dependencies with cache mount for faster rebuilds
 RUN pip install --upgrade pip && \
     pip install --timeout=7200 \
     terratorch[vllm]==1.2.1 \
-    # "vllm>=0.6.0" \
     'diffusers<=0.34.0' \
     ipython \
     nbformat \
@@ -121,8 +122,11 @@ RUN chmod +x /app/entrypoint.sh /app/healthcheck.sh
 
 # Switch to non-root user
 USER appuser
+
 WORKDIR /opt/app-root/src
+
 EXPOSE 8000
+
 HEALTHCHECK --interval=30s --timeout=10s --start-period=120s --retries=3 \
     CMD ["/bin/bash", "/app/healthcheck.sh"]
 

pipelines/entrypoint.sh

@@ -1,12 +1,12 @@
 #!/bin/bash
 set -e
 
-if [ "$APP_MODE" = "v" ]; then
-    echo "Starting in Standard TerraTorch mode..."
+if [ "$APP_MODE" = "vllm" ]; then
+    echo "Starting in VLLM mode"
 
     MODEL_NAME=${MODEL_NAME:-"ibm-nasa-geospatial/Prithvi-EO-2.0-300M-TL-Sen1Floods11"}
     PLUGIN_NAME=${PLUGIN_NAME:-"terratorch_segmentation"}
-    HF_HOME=${HF_HOME:-"/workspace/models/huggingface_cache"}
+    HF_HOME=${HF_HOME:-"/app/models/huggingface_cache"}
 
     echo "=========================================="
     echo "vLLM Prithvi Flood Detection Server"
@@ -18,8 +18,8 @@ if [ "$APP_MODE" = "v" ]; then
     echo ""
     # 1. Clear Hugging Face lock files
     # This prevents the 'OSError: PermissionError' if a previous pod crashed
-    echo "Checking for stale lock files in /workspace/models..."
-    find /workspace/models -name "*.lock" -delete
+    echo "Checking for stale lock files in /app/models..."
+    find /app/models -name "*.lock" -delete
 
     # 2. Verify GPU availability
     # This logs the GPU status to help debug 'Pending' or 'Runtime' errors