search

tensorflow / addons

Useful extra functionality for TensorFlow 2.x maintained by SIG-addons
Apache License 2.0
1.69k stars 611 forks source link

Build GPU custom Ops for windows #784

Closed seanpmorgan closed 1 year ago

seanpmorgan commented 4 years ago

Currently we only compile/use the CPU version of custom ops in Addons for windows. Users should still be able to leverage GPU ops from TF-Core but this is something we want to build/enable.

I don't think this is a huge lift since the gpu toolchain works on nix and windows, but needs some testing.

allen108108 commented 4 years ago

So, we can't use tensorflow-addons in Win10 system now ,right ?

seanpmorgan commented 4 years ago

So, we can't use tensorflow-addons in Win10 system now ,right ?

You can compile from source. See the README. It will only support CPU addons ops (but all GPU ops have a CPU fall back).

Alternatively we'll have a 0.7 release in the very near future to pip install.

allen108108 commented 4 years ago

Will you release the GPU version in future ?

seanpmorgan commented 4 years ago

Yes that is what this issue is here to track.

Just to be clear, prior to this issue being completed the addons package will still utilize GPU for underlying TF ops (matmul, etc.). But there will not be GPU implementations for the custom CUDA kernels in this repository... though they'll have CPU fallback.

allen108108 commented 4 years ago

Thanks for answering ~~~~

seanpmorgan commented 4 years ago

This has been unblocked: https://github.com/tensorflow/custom-op/pull/43#event-2902430632

seanpmorgan commented 4 years ago

Given the recent news: https://devblogs.microsoft.com/directx/directx-heart-linux/ Is this feature still requested from the community?

seanpmorgan commented 1 year ago

TensorFlow Addons is transitioning to a minimal maintenance and release mode. New features will not be added to this repository. For more information, please see our public messaging on this decision: TensorFlow Addons Wind Down

Please consider sending feature requests / contributions to other repositories in the TF community with a similar charters to TFA: Keras Keras-CV Keras-NLP