search

NVIDIA / apex

A PyTorch Extension: Tools for easy mixed precision and distributed training in Pytorch
BSD 3-Clause "New" or "Revised" License
8.42k stars 1.4k forks source link

is it possible to update `conda-forge/nvidia-apex` to a recent tag? #1730

Closed stas00 closed 1 year ago

stas00 commented 1 year ago

Describe the Bug

I see that the 22.03 apex conda-forge packages released are using a soon to be 2 year old apex: https://anaconda.org/conda-forge/nvidia-apex/files yet, someone rebuilds them every 3 months or so. Any reason not to update https://github.com/conda-forge/nvidia-apex-feedstock/blob/main/recipe/meta.yaml#L1 to 23.08 and continue updating it with new releases?

I realize that these tag versions correspond to the nvidia pytorch docker image versions, e.g. nvcr.io/nvidia/pytorch:23.08-py3

Thank you.

ptrblck commented 1 year ago

We did not create the conda-forge package and do not maintain it. I also don't know how it was built and which GPUs are supported so you might want to ask someone at conda-forge.

stas00 commented 1 year ago

Thank you for clarifying, @ptrblck!