Update default Binaries

facebookresearch / vissl

VISSL is FAIR's library of extensible, modular and scalable components for SOTA Self-Supervised Learning with images.

https://vissl.ai

MIT License

3.26k stars 334 forks source link

Update default Binaries #435

Closed amitbcp closed 3 years ago

amitbcp commented 3 years ago

Hi Team,

It seems the last pre-built binary was cut in March and since then a lot of things have been added and fixed in the repo. Maybe we can have a new pre-built binary.

Also all the installation doc says to used Cuda 10, any particular reason or should it work smoothly for coda 11 its respective pytorch version ?

prigoyal commented 3 years ago

Hi @amitbcp , thank you!

yes, @iseessel is driving the work on upgrading the packages.
it should work with CUDA 11. Please let us know if you encounter any issues.

iseessel commented 3 years ago

Hi @amitbcp yes we should support all versions of cuda that are compatible with pytorch. Note that we do not yet fully support Pytorch 1.8.1, 1.9.0, or 1.9.1.

Support for these should be coming early next week at the latest. A new minor release of vissl, accompanied by new prebuilt binaries should be coming by the end of this month.

iseessel commented 3 years ago

Hi @amitbcp we have released a new version of VISSL.

See our release notes for how to update: https://github.com/facebookresearch/vissl/releases/tag/v0.1.6.