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 2021/03/04 08:55:40 UTC
[GitHub] [tvm] huochaitiantang opened a new pull request #7586: [Relay] Fix relay op strategy for cuda dense int8
huochaitiantang opened a new pull request #7586:
URL: https://github.com/apache/tvm/pull/7586
The topi schedule `dense_int8.cuda` uses dp4a, whose types are int8 * int8 = int32.
This PR could ensure that, the quantized dense ops can find the more efficient schedule `dense_int8.cuda`, instead of `dense_small_batch.cuda` or `dense_large_batch.cuda`.
----------------------------------------------------------------
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] huochaitiantang commented on pull request #7586: [Relay] Fix relay op strategy for cuda dense int8
Posted by GitBox <gi...@apache.org>.
huochaitiantang commented on pull request #7586:
URL: https://github.com/apache/tvm/pull/7586#issuecomment-791169812
It seems that the schedule `dense_int8.cuda` does not support the `uint8` dtype, so I remove it.
And I add an autotvm task extraction test case, which tests relay graphs that contain int8 dense op.
----------------------------------------------------------------
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] icemelon9 merged pull request #7586: [Relay] Fix relay op strategy for cuda dense int8
Posted by GitBox <gi...@apache.org>.
icemelon9 merged pull request #7586:
URL: https://github.com/apache/tvm/pull/7586
----------------------------------------------------------------
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] icemelon9 commented on pull request #7586: [Relay] Fix relay op strategy for cuda dense int8
Posted by GitBox <gi...@apache.org>.
icemelon9 commented on pull request #7586:
URL: https://github.com/apache/tvm/pull/7586#issuecomment-792990624
Thanks @huochaitiantang
----------------------------------------------------------------
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