MPS test jobs are failing

[NicolasHug]

with

Traceback (most recent call last):
  File "/Users/runner/work/vision/vision/pytorch/vision/test/smoke_test.py", line 103, in <module>
    main()
  File "/Users/runner/work/vision/vision/pytorch/vision/test/smoke_test.py", line 99, in main
    smoke_test_torchvision_resnet50_classify("mps")
  File "/Users/runner/work/vision/vision/pytorch/vision/test/smoke_test.py", line 52, in smoke_test_torchvision_resnet50_classify
    img = read_image(str(SCRIPT_DIR / ".." / "gallery" / "assets" / "dog2.jpg")).to(device)
RuntimeError: MPS backend out of memory (MPS allocated: 0 bytes, other allocations: 0 bytes, max allowed: 1.70 GB). Tried to allocate 0 bytes on private pool. Use PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 to disable upper limit for memory allocations (may cause system failure).

example: https://github.com/pytorch/vision/actions/runs/9157662717/job/25174485539?pr=7990

[NicolasHug]

Seems that all the failures are related to MPS - pinging @qqaatw as you've been our POC on this so far :)
Do you have any idea what might be causing this? Has there been any change on the PyTorch core side related to MPS that could be the source? Thanks!

[qqaatw]

Hmm, it looks related to MPSAllocator. cc @kulinseth

[NicolasHug]

Looks like it wasn't directly related to MPS but rather related to the recent drop of macos x86 support from pytorch. After #8446 I see the MPS tests passing, so I'll close this issue.

Thank you for your prompt response @qqaatw !