You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2021/05/26 08:54:52 UTC
[GitHub] [tvm] jcf94 opened a new pull request #8138: [FastMath] Add fast_softmax support in fast_math pass
jcf94 opened a new pull request #8138:
URL: https://github.com/apache/tvm/pull/8138
Add fast_softmax support in fast_math pass.
p.s. AutoScheduler can easily support the fast_softmax implementation.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] comaniac commented on pull request #8138: [FastMath] Add fast_softmax support in fast_math pass
Posted by GitBox <gi...@apache.org>.
comaniac commented on pull request #8138:
URL: https://github.com/apache/tvm/pull/8138#issuecomment-848918375
Thanks @jcf94
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] merrymercy commented on pull request #8138: [FastMath] Add fast_softmax support in fast_math pass
Posted by GitBox <gi...@apache.org>.
merrymercy commented on pull request #8138:
URL: https://github.com/apache/tvm/pull/8138#issuecomment-849603578
This PR will break other code that does not use auto-scheduler.
If some existing scripts do not use auto-scheduler but enable FastMath pass, their performance will be degraded after this PR (for cpu backends), or they just cannot run (for GPU backends)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] jcf94 commented on pull request #8138: [FastMath] Add fast_softmax support in fast_math pass
Posted by GitBox <gi...@apache.org>.
jcf94 commented on pull request #8138:
URL: https://github.com/apache/tvm/pull/8138#issuecomment-849624653
> This PR will break other code that does not use auto-scheduler.
> If some existing scripts do not use auto-scheduler but enable FastMath pass, their performance will be degraded after this PR (for cpu backends), or they just cannot run (for GPU backends)
Yeah, actually I'm now writing schedules for fast_softmax in cuda, this does comes to be a problem.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] comaniac merged pull request #8138: [FastMath] Add fast_softmax support in fast_math pass
Posted by GitBox <gi...@apache.org>.
comaniac merged pull request #8138:
URL: https://github.com/apache/tvm/pull/8138
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org