GPU documentation is not clear

[cookielee77]

Looks like "device=gpu" only supports single gpu training but "device=cuda" can support multi-gpus training. In addition, "cuda" looks like only support double-precision which is memory consuming.

Is it possible to update the documentation for GPU argument more clearer (it took me a lot of time to find these secrets)? Also, is there a plan to support the single-prevision calculation with CUDA version in the near future? Double-precision is super memory-costly.

Thanks!

[StrikerRUS]

@cookielee77 Thanks a lot for bringing this up!

Is it possible to update the documentation for GPU argument more clearer (it took me a lot of time to find these secrets)?

Sure! Could you please suggest how these wordings can be improved to be clearer?
https://lightgbm.readthedocs.io/en/latest/Parameters.html#num_gpu
https://lightgbm.readthedocs.io/en/latest/Parameters.html#gpu_use_dp

Also, is there a plan to support the single-prevision calculation with CUDA version in the near future?

You can subscribe to the following GitHub issue to get all news about single-prevision mode with CUDA version.

#3836.

For now, no one have taken this feature request, unfortunately.

[shiyu1994]

@cookielee77 Yes, we have a plan to support single-precision calculation with CUDA in the very near future.

[BinchaoPeng]

hi，I would like to realize the difference between gpu version and cuda version, do you know where the doc is? @cookielee77 @jameslamb @StrikerRUS @shiyu1994 @cbecker

[StrikerRUS]

@BinchaoPeng Hi!
Sure, here it is: https://lightgbm.readthedocs.io/en/latest/Installation-Guide.html#build-cuda-version-experimental.

CUDA version is a re-written in CUDA language GPU version for systems where OpenCL is not available.

[BinchaoPeng]

@BinchaoPeng Hi!
Sure, here it is: https://lightgbm.readthedocs.io/en/latest/Installation-Guide.html#build-cuda-version-experimental.

CUDA version is a re-written in CUDA language GPU version for systems where OpenCL is not available.

okay，thank you！

[no-response]

This issue has been automatically closed because it has been awaiting a response for too long. When you have time to to work with the maintainers to resolve this issue, please post a new comment and it will be re-opened. If the issue has been locked for editing by the time you return to it, please open a new issue and reference this one. Thank you for taking the time to improve LightGBM!

[github-actions]

This issue has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this.