search

pythops / jetson-image

💾 Create minimalist, Ubuntu based images for the Nvidia jetson boards
372 stars 108 forks source link

Minimum image with CUDA support #132

Closed mpaszkuta closed 5 months ago

mpaszkuta commented 5 months ago

First of all, I would like to thank you for this project. It helps me save space on the SD card and time during development. I aim to create an image with the minimum CUDA tools installed for PyTorch Python development. There is a special version of PyTorch for Jetson modules that requires CUDA. Normally, these tools are installed with JetPack. The CUDA tools can be installed after the L4T integration. I have prepared a list of the minimum packages that need to be installed, but it is not possible to add this to Containerfile.rootfs.22_04 during the rootfs build. However, this should be realized in Containerfile.rootfs.22_04 after new sources are added. It would be beneficial to add the possibility of installing selected packages from JetPack, such as CUDA, OpenCV, etc.

pythops commented 5 months ago

Here is the PR to solve this issue https://github.com/pythops/jetson-image/pull/133 Just add the packages you want in the file l4t_packages.txt and build following the same instructions, I have tested it locally and it seems to be working well. Please give it a try and let me know if this approach suits you

pythops commented 5 months ago

I would like to merge the PR. please take a moment and give it a try. Any feedback is welcome :)

pythops commented 5 months ago

Merged