unskip test_nn conv2d groups tests

[amathews-amd]

🐛 Describe the bug

test_nn	test_Conv2d_groups_nobias	(main.TestNN)
test_nn	test_Conv2d_groups_nobias_v2	(main.TestNN)
test_nn	test_softmax_cuda_float16	(main.TestNNDeviceTypeCUDA)
test_nn	test_softmax_cuda_float32	(main.TestNNDeviceTypeCUDA)
test_nn	test_softmax_results_cuda_float16	(main.TestNNDeviceTypeCUDA)
test_nn	test_softmax_results_cuda_float32	(main.TestNNDeviceTypeCUDA)
test_nn	test_upsamplingNearest2d_launch_fail_cuda	(main.TestNNDeviceTypeCUDA)

Versions

ROCm version: 4.5.2 PyTorch commit: https://github.com/pytorch/pytorch/commit/4b4f652f799c732a7280052e44a803a4cf6084de March 17, 2022 Test Audit logs ROCm Test1: https://ossci-raw-job-status.s3.amazonaws.com/log/5590875721 ROCm Test2: https://ossci-raw-job-status.s3.amazonaws.com/log/5590875831 ROCm distributed tests: https://ossci-raw-job-status.s3.amazonaws.com/log/5590691830 CUDA Test1: https://ossci-raw-job-status.s3.amazonaws.com/log/5590724129 CUDA Test2: https://ossci-raw-job-status.s3.amazonaws.com/log/5590724019 CUDA distributed tests: https://ossci-raw-job-status.s3.amazonaws.com/log/5590723920