Skip to content

Skip the failing unit tests from the FusedRMSNorm PR #85

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jithunnair-amd merged 2 commits into from
Aug 8, 2022
Merged

Skip the failing unit tests from the FusedRMSNorm PR #85

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
32f70d6
Skip the failing unit tests from the FusedRMSNorm PR
Aug 5, 2022
5493703
Update test_lamb.py
jithunnair-amd Aug 8, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions tests/L0/run_optimizers/test_fused_optimizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,13 +94,15 @@ def __init__(self, *args, **kwargs):
self.ref_optim = torch.optim.Adam
self.fused_optim = apex.optimizers.FusedAdam

@unittest.skip("Skipped the test since a regression introduced from PyTorch upstream: due to https://github.com/pytorch/pytorch/issues/80809#issuecomment-1175211598. Please also refer to https://github.com/ROCmSoftwarePlatform/apex/issues/82")
def test_float(self):
self.gen_single_type_test(param_type=torch.float)

@unittest.skip("NaN issue observed on ROCm as of 12/1/2021. The failing unit test is introduced by a PyTorch commit sometime in between rocm/pytorch:rocm4.3.1_ubuntu18.04_py3.6_pytorch_1.9.0 and 2021/12/01. Please refer to https://github.com/ROCmSoftwarePlatform/apex/issues/63")
def test_half(self):
self.gen_single_type_test(param_type=torch.float16)

@unittest.skip("Skipped the test since a regression introduced from PyTorch upstream: due to https://github.com/pytorch/pytorch/issues/80809#issuecomment-1175211598. Please also refer to https://github.com/ROCmSoftwarePlatform/apex/issues/82")
@unittest.skipIf(torch.cuda.device_count()<2, "more than 1 GPU required")
def test_multi_device(self):
devices = ("cuda:0", "cuda:1")
Expand Down Expand Up @@ -167,6 +169,7 @@ def test_fp16_output(self):
self.assertLessEqual(max_abs_diff, self.max_abs_diff)
self.assertLessEqual(max_rel_diff, self.max_rel_diff)

@unittest.skip("Skipped the test since a regression introduced from PyTorch upstream: due to https://github.com/pytorch/pytorch/issues/80809#issuecomment-1175211598. Please also refer to https://github.com/ROCmSoftwarePlatform/apex/issues/82")
def test_adam_option(self):
nelem = 1
adam_option = {'lr':0.01, 'betas':(0.6, 0.9), 'eps':3e-06,
Expand Down
1 change: 1 addition & 0 deletions tests/L0/run_optimizers/test_lamb.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -285,6 +285,7 @@ def test_float(self):
def test_half(self):
self.gen_single_type_test(param_type=torch.float16)

@unittest.skip("Skipped the test since it failed the accuracy test on the PyTorch as of 8/1/2022. Please refer to https://github.com/ROCmSoftwarePlatform/apex/issues/83")
@unittest.skipIf(torch.cuda.device_count()<2, "more than 1 GPU required")
def test_multi_device(self):
devices = ("cuda:0", "cuda:1")
Expand Down