PR #6461 rebase - SemanticDiff

rebase #6461

ManfeiBai merged 24 commits into ManfeiBai-patch-62 from master

Disambiguate `benchmarks` directory search. (#6434)

599b9256

Revert "Disambiguate `benchmarks` directory search. (#6434)" (#6440)

bc4fee20

Fix model skip yaml file not found issue (#6428)

e2012881

Check precision only if on CUDA and not inductor. (#6407)

30591ad8

Enable CUDA support for PyTorch on CI. (#6070)

d5e1b49e

Move tests that need CUDA to `test_operations.py`. (#6160)

16b6bd1b

Re-land "aggregate.py: add tab format (#6405)", reverted in #6424 (#6…

904676be

Add some helpful debugging notes to TROUBLESHOOTING docs (#6430)

17161eb6

Properly lower aten::lerp (#6339)

d73bd833

Remove `TPU_C_API` device type (#6435)

d686d65b

Enable install `google-cloud-sdk-gke-gcloud-auth-plugin` to enable TP…

479559c4

aggregate.py: fix plotting of more than 3 backends (#6449)

800a3254

nightly.sh: also run with non-dynamo (i.e. lazy tensor) XLA (#6432)

c154b991

Re-instate the missing CPU CI tests and their coverage reports (#6448)

52ca01e7

Add `benchmarks/dynamo` search option for finding PyTorch skip file. …

f4849154

follow-up for #6420 (#6450)

89f7f43b

Disable TestPoolingNNDeviceTypeXLA temporarily. (#6452)

d0a5c34e

Fix benchmark process terminaion issue (#6441)

d1a454fe

Disable test_EmbeddingBag_per_sample_weights_and_new_offsets (#6456)

21653dc9

Unify TORCH_LIBRARY definitions (#6455)

19923bd1

torchbench_model.py: fix YAML skip file path (#6457)

510bb8ea

[benchmarks] Add support for AMP execution. (#6447)

ad822bf0

Force guards when performing boolean operations on XLASymNode (#6433)

fe604e97