how to uninstall CUDA10.0 in JP4.2 Nano

I want to use a low version of tensorflow and want to uninstall the CUDA10.0.
How to do it?
Many thanks

The CUDA version is tied to other components of the install. It won’t work to install a different version from a different release. You’d need an older JetPack/SDK Manager release with a full flash, but R32.1 (JetPack/SDKM 4.2) is the first release for Nano.

I meet two problems.

One is that I want to install JP4.2 from SDKmanger and I don’t want to install the CUDA10.0. When I do not choose the SDK component, the other SDK such as computer vision would uninstall at the same time but in fact I want to install the computer vision and don’t want to install the cuda10.0. How to solve it?

Another is that I try to install R32.1 from https://developer.nvidia.com/embedded/linux-tegra without other SDK component but I see I just use about 16g micro SD card (actually the micro SD card is about 32g,but the system just use 16g).How to solve it?

Many thanks!

Unfortunately SDK Manager changed to allow only an all or none install of “extras”. However, CUDA is required for most (all?) of those other components…they won’t function without this dependency. Computer vision depends on this if you wish to use the NVIDIA hardware.

I’ve seen issues reported for certain size SD cards, but don’t know the details. Someone else may be able to offer steps to use a 32GB SD card.