viable/strict/1789642231: [ROCm] Unskip assorted GPU tests and fix two ROCm arch skips (#197340)
- PyTorch: 1095 events in the last 90 days
- PyTorch: 1082th Release in the last 90 days
- Previous: earlier the same day · trunk/58c023503c29f9553198c45a2d38ff352e930211: Unify LazyLevelZero stub macro into a single ZE_STUB (#197209)
What happened
Removes ROCm skips that no longer correspond to a failure: five ShardedTensor tests that do not use RPC (the skip was carried over from the RPC-based tests when #152583 added them), test_float8_scale_result, the semi-structured cutlass_mm functionalization decomposition test (FakeTensorMode only, no CUTLASS kernel), test_var_mean and test_cpu_scalar_with_gpu_tensor_cpp in test_torchinductor, and the in-body ROCm ear…
Summary assembled by rule from the sources below