Skip to content

Conversation

xinyazhang
Copy link

@xinyazhang xinyazhang commented Aug 18, 2023

This is due to incorrect atol/rtol settings for torch.half and torch.bfloat16 data types.

Fixes #SWDEV-416500

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.
@xinyazhang xinyazhang force-pushed the xinyazhang/fix-test_Conv2d_groups branch from 8e038f9 to b47b956 Compare August 18, 2023 06:45
@pruthvistony
Copy link
Collaborator

The change is generic, doesnt affect CUDA is it?

@xinyazhang
Copy link
Author

This change is generic and also applies to CUDA. However the question about CUDA is how can it achieve 1e-5 precision with float16 or bfloat16 when validating the output.

Copy link
Collaborator

@jithunnair-amd jithunnair-amd left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@xinyazhang Let's still gather the MIOpenDriver configs so we can file an MIOpen ticket. Whether it gets fixed in ROCm5.7 is up to them and Release Management. Meanwhile, merging this PR as a workaround.

@jithunnair-amd jithunnair-amd merged commit e9b273d into rocm5.7_internal_testing Aug 18, 2023
pruthvistony pushed a commit that referenced this pull request Sep 11, 2023
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32
dnikolaev-amd pushed a commit that referenced this pull request Sep 11, 2023
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32
jithunnair-amd pushed a commit that referenced this pull request Sep 12, 2023
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32

Co-authored-by: Xinya Zhang <[email protected]>
dnikolaev-amd pushed a commit that referenced this pull request Sep 21, 2023
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32
pruthvistony pushed a commit that referenced this pull request Dec 6, 2023
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32
dnikolaev-amd pushed a commit that referenced this pull request Dec 22, 2023
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32
dnikolaev-amd pushed a commit that referenced this pull request Jan 9, 2024
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32
pruthvistony pushed a commit that referenced this pull request Jan 10, 2024
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32

Co-authored-by: Xinya Zhang <[email protected]>
pruthvistony pushed a commit that referenced this pull request Jan 21, 2024
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32
pruthvistony pushed a commit that referenced this pull request Jan 22, 2024
* Fix test_Conv2d_groups related errors (SWDEV-416489)

This is due to incorrect atol/rtol settings for torch.half and
torch.bfloat16 data types.

* More adjustment on Navi 32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants