-
Notifications
You must be signed in to change notification settings - Fork 24.7k
[MPS] Fix crash when mm is invoked with mixed dtypes #143948
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/143948
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 1fd7ded with merge base e1abbe1 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
Simply by copy-n-pasting check from https://github.com/pytorch/pytorch/blob/a7915c56f6a62266490be355b3d823b1e447a475/aten/src/ATen/native/cuda/Blas.cpp#L254-L257 [ghstack-poisoned]
Simply by copy-n-pasting check from https://github.com/pytorch/pytorch/blob/a7915c56f6a62266490be355b3d823b1e447a475/aten/src/ATen/native/cuda/Blas.cpp#L254-L257 [ghstack-poisoned]
@pytorchbot merge -f "no need to run a trunk testing" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
TODO: as followup PR replace this weird logic with shaders Fixes #143932 Pull Request resolved: #143949 Approved by: https://github.com/Skylion007 ghstack dependencies: #143948
Alas, PythonPrinter would not work here, not would CppPrinter, so start building MetalPrinter. `pytest test/inductor/test_torchinductor.py -k _mps` score is 474 failed, 277 passed, 32 skipped Before this change: `pytest test/inductor/test_torchinductor.py -k _mps` reported 506 failed, 245 passed, 32 skipped Pull Request resolved: #143973 Approved by: https://github.com/jansel ghstack dependencies: #143948, #143949
By calling `metal::min` and `metal::max` respectively with argument typecast to a common type to avoid ambiguous calls errors TODO: Implement NaN propagation for both eager and compile, see #143976 `pytest test/inductor/test_torchinductor.py -k _mps` score is 460 failed, 291 passed, 32 skipped Pull Request resolved: #143977 Approved by: https://github.com/jansel ghstack dependencies: #143948, #143949, #143973
At the moment by generating multiple MetalLibraries `pytest test/inductor/test_torchinductor.py -k _mps` score is 434 failed, 317 passed, 32 skipped Pull Request resolved: #143998 Approved by: https://github.com/jansel, https://github.com/ruidazeng ghstack dependencies: #143948, #143949, #143973, #143977
Stack from ghstack (oldest at bottom):
torch.add(x,y, alpha=2)
crash #143949Simply by copy-n-pasting check from
pytorch/aten/src/ATen/native/cuda/Blas.cpp
Lines 254 to 257 in a7915c5