ciflow/xpu/193295: [inductor] Fix round-11 test config leakage and gradient tolerance
- PyTorch: 2130 events in the last 90 days
- PyTorch: 2101th Release in the last 90 days
- Previous: earlier the same day · ciflow/xpu/189675
What happened
Round-11 CI reported two failures in the test additions. test_batch_linear_lhs_skipped_under_default_config failed on GPU runners with "The length of the sequences mismatch: 3 != 1". The guard options were applied with a class-level torch._inductor.config.patch on _BatchLinearLHSGuardConfigBase, but instantiate_device_type_tests invokes setUpClass while creating the CUDA/XPU variants because get_primary_device() rai…
Summary assembled by rule from the sources below