Enabled TEGroupedMLP test.

[wenchenvincent]

No description provided.

[gurpreet-dhami]

@wenchenvincent , @lcskrishna
Do we want to remove the original markers e.g. "internal" ?
At some point I wanted to run all internal tests as well and mark the failing ones for further debugging ?
I suggest to keep the original markers as it is. And add additional markers ( e.g. test_on_rocm ) for the tests like this ?
It is easier for subsequent IFU updates as well.

[wenchenvincent]

@wenchenvincent , @lcskrishna Do we want to remove the original markers e.g. "internal" ? At some point I wanted to run all internal tests as well and mark the failing ones for further debugging ? I suggest to keep the original markers as it is. And add additional markers ( e.g. test_on_rocm ) for the tests like this ? It is easier for subsequent IFU updates as well.

@gurpreet-dhami We should keep most of the internal markers. For this one, I wanted to enable the specific tests for TEGroupedMLP to make sure that we test it. I like your idea of adding additional markers instead of removing the internal markers. I will make the change accordingly.

[wenchenvincent]

@gurpreet-dhami @lcskrishna I refactored the test markers. Could you please review the change?

[gurpreet-dhami]

@wenchenvincent : Change looks good to me. I hope the tests are passing for you.
I saw that CI failed on another unrelated test http://ml-ci-internal.amd.com:8080/job/Megatron-LM/job/Megatron-LM/view/change-requests/job/PR-22/3/console
tests/unit_tests/transformer/test_transformer_block.py tests/unit_tests/transformer/test_transformer_block.py ........................FFFFFFFF........Fatal Python error: Aborted

I ran again. lets see how it goes.

[lcskrishna]

@wenchenvincent : Change looks good to me. I hope the tests are passing for you. I saw that CI failed on another unrelated test http://ml-ci-internal.amd.com:8080/job/Megatron-LM/job/Megatron-LM/view/change-requests/job/PR-22/3/console tests/unit_tests/transformer/test_transformer_block.py tests/unit_tests/transformer/test_transformer_block.py ........................FFFFFFFF........Fatal Python error: Aborted

I ran again. lets see how it goes.

Seems like UTs are failing with this one in another PR as well. @gurpreet-dhami Can you take a look at this one.

[gurpreet-dhami]

For some reason, CI is failing on your PR on transformer_block tests.

It is passing on rocm_dev branch

passing on rocm_dev - http://ml-ci-internal.amd.com:8080/job/Megatron-LM/job/Megatron-LM/job/rocm_dev/19/

failing on TEgrouped gemm PR: http://ml-ci-internal.amd.com:8080/job/Megatron-LM/job/Megatron-LM/view/change-requests/job/PR-22/4/