The l4t-base:r32.2.1 base image (jetson) provided by NVIDIA does not have cuda binaries Pytorch is looking for. I am unable to download source files from Jetpack 4.2.2 (specifically cuda-l4t-repo-10.x.x-arm64.deb) to include within docker build.
I am building a docker container for jetsons. I used the l4t image from NVIDIA, but it is missing some binaries pytorch is looking for (I understand you're supposed to mount some files for CUDA, but I don't want to). I therefore was going to install CUDA 10 directly into the container.
I am trying to download the source files for Jetpack 4.2.2. From what I have read, there should be a file, "cuda-repo-l4t-10-0-local-_arm64.deb". However, I only see cuda-repo-cross-aarch64-10-0-local-10.0.326_1.0-1_all.deb and cuda-repo-ubuntu1804-10-0-local-10.0.326-410.108_1.0-1_amd64.deb.
I created an Ubuntu 18 container to run the SDK in. I run the below command and get the following files... 2019_06_23_2352-26660552-NVIDIA_Nomad_2019.2.19174.2352_Release_External-L4T.linux.deb, libopencv-dev_3.3.1-2-gb3f86dcd5_amd64.deb, Jetson_Linux_R32.2.1_aarch64.tbz2, libopencv-python_3.3.1-2-gb3f86dcd5_amd64.deb NVIDIA_VisionWorks_References.zip, libopencv-samples_3.3.1-2-gb3f86dcd5_amd64.deb NsightSystems-linux-public-2019.4.1.10-a76094a.deb libopencv_3.3.1-2-gb3f86dcd5_amd64.deb Tegra_Linux_Sample-Root-Filesystem_R32.2.1_aarch64.tbz2 libvisionworks-repo_1.6.0.500n_amd64.deb cuda-repo-cross-aarch64-10-0-local-10.0.326_1.0-1_all.deb libvisionworks-sfm-repo_0.90.4_amd64.deb cuda-repo-ubuntu1804-10-0-local-10.0.326-410.108_1.0-1_amd64.deb libvisionworks-tracking-repo_0.88.2_amd64.deb devtools_docs.zip sdkml3_jetpack_l4t_422_rev1_b30.json
sdkmanager --cli downloadonly --user $NV_USER \
--view live --host --logintype devzone --product Jetson --version 4.2 \
--targetos Linux --target $DEVICE_ID --flash all --license accept \
--downloadfolder /tmp/cuda_pkg
Desired results:
- download jetson jetpack ARM specific cuda packages during docker build
- all cuda binaries are available in container
instead of:
root@24081bcd996b:/usr/local/cuda-10.0/lib64# ls
libcudadevrt.a libcudart_static.a stubs
should look more like:
libOpenCL.so libcufft.so.10.0 libnppc.so libnppif_static.a libnpps.so.10.0.176
libOpenCL.so.1 libcufft.so.10.0.176 libnppc.so.10.0 libnppig.so libnpps_static.a
libOpenCL.so.1.0 libcufft_static.a libnppc.so.10.0.176 libnppig.so.10.0 libnvToolsExt.so
libOpenCL.so.1.0.0 libcufftw.so libnppc_static.a libnppig.so.10.0.176 libnvToolsExt.so.1
libaccinj64.so libcufftw.so.10.0 libnppial.so libnppig_static.a libnvToolsExt.so.1.0.0
libaccinj64.so.10.0 libcufftw.so.10.0.176 libnppial.so.10.0 libnppim.so libnvblas.so
libaccinj64.so.10.0.176 libcufftw_static.a libnppial.so.10.0.176 libnppim.so.10.0 libnvblas.so.10.0
libcublas.so libcuinj64.so libnppial_static.a libnppim.so.10.0.176 libnvblas.so.10.0.176
libcublas.so.10.0 libcuinj64.so.10.0 libnppicc.so libnppim_static.a libnvblas.so.10.0.480
libcublas.so.10.0.176 libcuinj64.so.10.0.176 libnppicc.so.10.0 libnppist.so libnvgraph.so
libcublas.so.10.0.480 libculibos.a libnppicc.so.10.0.176 libnppist.so.10.0 libnvgraph.so.10.0
libcublas_device.a libcurand.so libnppicc_static.a libnppist.so.10.0.176 libnvgraph.so.10.0.176
libcublas_static.a libcurand.so.10.0 libnppicom.so libnppist_static.a libnvgraph_static.a
libcudadevrt.a libcurand.so.10.0.176 libnppicom.so.10.0 libnppisu.so libnvrtc-builtins.so
libcudart.so libcurand_static.a libnppicom.so.10.0.176 libnppisu.so.10.0 libnvrtc-builtins.so.10.0
libcudart.so.10.0 libcusolver.so libnppicom_static.a libnppisu.so.10.0.176 libnvrtc-builtins.so.10.0.176
libcudart.so.10.0.176 libcusolver.so.10.0 libnppidei.so libnppisu_static.a libnvrtc.so
libcudart_static.a libcusolver.so.10.0.176 libnppidei.so.10.0 libnppitc.so libnvrtc.so.10.0
libcudnn.so libcusolver_static.a libnppidei.so.10.0.176 libnppitc.so.10.0 libnvrtc.so.10.0.176
libcudnn.so.7 libcusparse.so libnppidei_static.a libnppitc.so.10.0.176 stubs
libcudnn.so.7.6.3 libcusparse.so.10.0 libnppif.so libnppitc_static.a
libcudnn_static.a libcusparse.so.10.0.176 libnppif.so.10.0 libnpps.so
libcufft.so libcusparse_static.a libnppif.so.10.0.176 libnpps.so.10.0
You can locate the url in the sdkml3_jetpack_l4t_422_rev1_b30.json to download the debain files that you are looking for .
Hope this helps