[Intel GPU] Customized XPU behaviour in indexing, group norm #134453
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/134453
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (4 Unrelated Failures)As of commit ddd74a3 with merge base d14fe3f ( BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
aten/src/ATen/native/group_norm.cpp
Outdated
| X.suggest_memory_format() : at::MemoryFormat::Contiguous; | ||
|
|
||
| TORCH_CHECK(X.is_contiguous(memory_format)); | ||
| TORCH_CHECK(X.is_contiguous(memory_format) || (X.device().is_xpu() && memory_format == at::MemoryFormat::Contiguous)); |
There was a problem hiding this comment.
X.device().is_xpu() && memory_format == at::MemoryFormat::Contiguous is always True. Because X.device().is_cpu() is False and then auto memory_format = at::MemoryFormat::Contiuous. So it means the statement is always True if X is an Intel GPU tensor. Is that expected?
There was a problem hiding this comment.
Yeah, I change the code to make sure we have contiguous input&output always, this should be more proper than the above codes.
ZhiweiYan-96
added
ciflow/xpu
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
…#134453) Pull Request resolved: pytorch#134453 Approved by: https://github.com/EikanWang, https://github.com/albanD ghstack dependencies: pytorch#133980
Stack from ghstack (oldest at bottom):