> Getting the actual GPU working was also painful, so I’ll leave the details here for future adventurers:

> # Data Center/HGX-Series/HGX H100/Linux aarch64/12.8 seem to work! wget https://us.download.nvidia.com/tesla/570.195.03/NVIDIA-Linux...

> ...

Nothing makes you feel more "I've been there" than typing inscrutable arcana to get a GPU working for ML work...