Nikita Shulga

@pytorchbot revert -m 'This breaks lint https://hud.pytorch.org/pytorch/pytorch/commit/7f256fff77c49729131aa6d092e60e891d0c4948' -c ignoredsignal

@cpuhrsch why do you think failures were spurious?

@anikethc at the very least, .py files should remain valid executable files, which is not the case after your changes, is it?

Assigning to myself to get a quick repro...

Hi-pri as in 2.3.0, it just crashes MPS runtime: ``` % python bug-124850.py STAGE:2024-04-24 11:27:36 16333:8699272 ActivityProfilerController.cpp:314] Completed Stage: Warm Up Error: command buffer exited with error status. The Metal...

IMO we should make the deprecation if and only if we achieve feature parity between newest and latest Python versions. Right now python-3.12 lacks torch.compile support, and if adding it...

Transferring to the torchdata project, though please note that it's not really maintained by anyone right now

But `centos:8` already EOLed in 2021, see https://www.centos.org/centos-linux-eol/ IMO we should migrate Linux docker base to https://github.com/pypa/manylinux / https://quay.io/repository/pypa/manylinux2014_x86_64?tab=tags&tag=latest

I have an even simpler 1-line reproducer: ``` % python -c "import torch;print(torch.ones(32, 4096, 4096, device='mps').shape)" /AppleInternal/Library/BuildRoots/5381bdfb-27e8-11ed-bdc1-96898e02b808/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShaders/MPSCore/Types/MPSNDArray.mm:705: failed assertion `[MPSTemporaryNDArray initWithDevice:descriptor:] Error: product of dimension sizes > 2**31' ```

And I can reproduce matmul crash using following pure-MPS script: ```swift import MetalPerformanceShadersGraph let graph = MPSGraph() let x = graph.constant(1, shape: [32, 4096, 40], dataType: .float32) let y =...