Reduce to single CUDA arch (7.5) and limit parallel jobs to reduce memory usage

Dockerfile

@@ -19,7 +19,9 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     && rm -rf /var/lib/apt/lists/*
 
 # Build FoundationPose C++ extensions (requires GPU)
+# MAX_JOBS=1 to reduce memory usage during compilation
 WORKDIR /app/FoundationPose
+ENV MAX_JOBS=1
 RUN cd bundlesdf/mycuda && pip install . --no-build-isolation
 RUN cd mycpp && python setup.py build_ext --inplace
 

Dockerfile.base

@@ -6,8 +6,8 @@ ENV CUDA_HOME=/usr/local/cuda
 ENV PATH=${CUDA_HOME}/bin:${PATH}
 ENV LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${LD_LIBRARY_PATH}
 
-# Only build for T4 (7.5) and A100 (8.0) - HuggingFace hardware
-ENV TORCH_CUDA_ARCH_LIST="7.5;8.0"
+# Only build for T4 (7.5) - reduces compilation memory by 50%
+ENV TORCH_CUDA_ARCH_LIST="7.5"
 
 # Install minimal runtime dependencies
 # Remove problematic CUDA repo and install packages