Blog post to address Pytorch channel deprecation #2381
Conversation
✅ Deploy Preview for conda-forge-previews ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
|
|
||
| ## Our Commitment | ||
|
|
||
| - Conda-forge is actively working with PyTorch maintainers and the broader open-source community to address potential gaps, ensure stability, and improve package availability. |
There was a problem hiding this comment.
I’m not sure that this would be my first commitment or any commitment at all on the list.
My only commitment is to high quality packaging and working in the open.
If and when those involved in the upstream pytorch package want to get involved in the conda-forge effort then I am committed to review it based on my availability as a volunteer.
There was a problem hiding this comment.
As the original author of the conda-forge package and an active maintainer of the feedstock, I’m not sure this blog post speaks for me.
To be honest, the announcement from PyTorch changes nothing on my efforts to package things at conda forge.
Conda forge has always been a place where people intereested in contributing high quality packages can come.
I would reiterate our shared value for high quality packaging and collaboration. Point to the blog post about transitioning from Anaconda’s default channel, but remove much of the discussion relating to the compatibility of our package other than what is supported by the latest version.
PyTorch has been notoriously difficult to package and many older versions are broken due to lack of testing.
Today, windows support is non existent. I don’t have a commitment to creating a package for it. Rather, it will come from my (and others) commitment to collaboration.
|
I do want to thank you for attempting to write something about this situation. It has been difficult for me to write something about Anacondas licensing enforcement and PyTorch’s deprecation of their conda package. However for me it is simply the status quo. If you are interested in package interactions and the c level and want the latest versions of packages, conda-forge may be a good option for you. others may have a different opinion about how we should phrase things. But. Thank you for getting this conversation going!!!! ❤️🙂 |
| mamba install pytorch-cpu | ||
| mamba install pytorch-gpu # for GPU support | ||
| # OR | ||
| conda install pytorch-cpu | ||
| conda install pytorch-gpu # for GPU support |
There was a problem hiding this comment.
Let's keep using the same as above.
| mamba install pytorch-cpu | |
| mamba install pytorch-gpu # for GPU support | |
| # OR | |
| conda install pytorch-cpu | |
| conda install pytorch-gpu # for GPU support | |
| mamba/conda install pytorch-cpu | |
| mamba/conda install pytorch-gpu # for GPU support |
PR Checklist:
docs/orcommunity/, you have added it to the sidebar in the corresponding_sidebar.jsonfileHey conda-forge team!
I put together a draft for a blog post to address and reassure conda-forge users about the recent news regarding the deprecation of the PyTorch channel (see the announcement).
I tried to cover the key points, but there might be some inaccuracies or assumptions in there, so I’d really appreciate it if you could take a look and share your feedback. @isuruf @hmaarrfk @jaimergp