Indeed we both have 4090's so that might explain the issue.
It sucks that nvidia is more buggy in general on linux because they're currently the only realistic* option for working with a bunch of ML stuff.
* Yes ROCM is gaining support, but it's still trailing behind CUDA unfortunately and you can't beat the 4090/3090 for ML in the consumer range right now.
It sucks that nvidia is more buggy in general on linux because they're currently the only realistic* option for working with a bunch of ML stuff.
* Yes ROCM is gaining support, but it's still trailing behind CUDA unfortunately and you can't beat the 4090/3090 for ML in the consumer range right now.