Hi @jubin0927, the JetPack 4 containers are incompatible with JetPack 5, so you can’t do that. On JetPack 4, many of the JetPack components like CUDA/cuDNN/TensorRT/GStreamer are mounted into the container dynamically from the host device. So on JetPack 5, those expected components wouldn’t be there. Instead please migrate to using nvcr.io/nvidia/l4t-base:r35.2.1