Base image with PyTorch 2.9/cuda 13?

AI & Data Science NVIDIA AI Workbench
, , ,
1

I was referred here from the Spark forums: https://forums.developer.nvidia.com/t/which-pytorch-base-image-to-use-in-ai-workbench-with-dgx-spark/350476/14

What is the quickest way to an AI Workbench base image with the latest PyTorch and CUDA with GB support? Do I need to fiddle with labels to convert the existing NGC docker container to a base image or is there a simpler way?

Thanks,

Simon.

1 Like
2

No resolution here?

3

Not for me, no. I abandoned AI Workbench, spent an evening making my own docker container based on the ngc image and now I’m using that.

The only slightly tedious part was fiddling with the pip dependencies so that I didn’t accidentally transitively bring in a CPU-only torch wheel, but so far so good 🤞.

Simon.

1 Like