Ray or hvd
Closed this issue · 2 comments
1tac11 commented
Hi there
Why is horovod needed if you have ray? Ray can also run on multiple GPUs. And do both not interfere? And where are the parameters handled? In rays database or hvd?
Kind regards
Denys88 commented
Hi, I see you already closed it. But Ill answer anyway.
Ray is used for cpu environments only. Also I didn't find the way how to use it in efficient way for multigpu use-case with NCCL for example.
Imaging you are going to use my library to train atari and you have multi gpu setup. You will use both libraries for this case.
For the IsaacGym case they provide their own vectorized env on gpu so horovod is the fastests.
Thanks,
Denys
1tac11 commented
Hi, thank you for answering. Ok, and horovod can also be used if spread
over multiple machines? Maybe that's not necessary for the GPU case.
Kind regards
Denys Makoviichuk ***@***.***> schrieb am Di., 18. Jan. 2022,
00:03:
… Hi, I see you already closed it. But Ill answer anyway.
Ray is used for cpu environments only. Also I didn't find the way how to
use it in efficient way for multigpu use-case with NCCL for example.
Imaging you are going to use my library to train atari and you have multi
gpu setup. You will use libraries for this case.
For the IsaacGym case they provide their own vectorized env on gpu so
horovod is the fastests.
—
Reply to this email directly, view it on GitHub
<#116 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AVX3ODBCQIJ2PSFL5BESWCDUWSN5FANCNFSM5MFQ6HPQ>
.
Triage notifications on the go with GitHub Mobile for iOS
<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675>
or Android
<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
You are receiving this because you modified the open/close state.Message
ID: ***@***.***>