From 3e07b25436445cbf4a1d5a39ea36acfa69ee646a Mon Sep 17 00:00:00 2001 From: akashveramd Date: Mon, 7 Jul 2025 14:54:34 -0700 Subject: [PATCH] Skipped *_stress_cuda UTs in test_c10d_gloo in release/2.5 branch. (#2317) In this PR, I have skipped *_stress_cuda UTs in test_c10d_gloo in release/2.5 branch. The tests are also skipped upstream. The test was failing for Jira ticket- https://ontrack-internal.amd.com/browse/SWDEV-503871 Tested using docker image- compute-artifactory.amd.com:5000/rocm-plus-docker/framework/compute-rocm-rel-6.4:114_ubuntu22.04_py3.10_pytorch_release-2.5_a1ad153 --- test/distributed/test_c10d_gloo.py | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/test/distributed/test_c10d_gloo.py b/test/distributed/test_c10d_gloo.py index ee1e04a2be792..c8ded6f3f9922 100644 --- a/test/distributed/test_c10d_gloo.py +++ b/test/distributed/test_c10d_gloo.py @@ -52,6 +52,7 @@ retry_on_connect_failures, run_tests, skip_but_pass_in_sandcastle, + skipIfRocm, TestCase, ) @@ -385,6 +386,7 @@ def test_broadcast_stress(self): inputs = [torch.tensor([i * self.world_size + self.rank]) for i in range(1000)] self._test_broadcast_stress(inputs) + @skipIfRocm @skip_if_lt_x_gpu(2) @requires_gloo() def test_broadcast_stress_cuda(self): @@ -490,6 +492,7 @@ def test_allreduce_stress(self): inputs = [torch.tensor([i + self.rank]) for i in range(1000)] self._test_allreduce_stress(inputs) + @skipIfRocm @skip_if_lt_x_gpu(2) @requires_gloo() def test_allreduce_stress_cuda(self): @@ -922,6 +925,8 @@ def test_scatter_stress(self): @skip_but_pass_in_sandcastle( "Test is flaky, see https://github.com/pytorch/pytorch/issues/15963" ) + + @skipIfRocm @skip_if_lt_x_gpu(2) @requires_gloo() def test_scatter_stress_cuda(self): @@ -1096,6 +1101,7 @@ def test_gather_stress(self): inputs = [torch.tensor([i + self.rank]) for i in range(1000)] self._test_gather_stress(inputs, lambda t: t.clone()) + @skipIfRocm @skip_if_lt_x_gpu(2) @requires_gloo() def test_gather_stress_cuda(self): @@ -1231,6 +1237,7 @@ def test_allgather_stress(self): inputs = [torch.tensor([i + self.rank]) for i in range(1000)] self._test_allgather_stress(inputs, lambda t: t.clone()) + @skipIfRocm @skip_if_lt_x_gpu(2) @requires_gloo() def test_allgather_stress_cuda(self): @@ -1417,6 +1424,7 @@ def test_reduce_stress(self): inputs = [torch.tensor([i + self.rank]) for i in range(1000)] self._test_reduce_stress(inputs) + @skipIfRocm @skip_if_lt_x_gpu(2) @requires_gloo() def test_reduce_stress_cuda(self):