Skip Metal-only gated delta kernel tests on other backends(fix CI) - #4522
Merged
Merged
Conversation
nastya236
approved these changes
Sep 16, 2026
| self.assertTrue(hf.dtype == mx.float32, msg="State " + msg) | ||
|
|
||
| @unittest.skipIf(not mx.is_available(mx.gpu), "No GPU available") | ||
| # The chunk size only selects between Metal kernels. Other backends always |
Collaborator
There was a problem hiding this comment.
I think the comment should be that we have fast implementation only on metal for now, it is independent from GATED_DELTA_CHUNK
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The
test_gated_delta_{sequential,simdgroup,nax}tests setGATED_DELTA_CHUNKto pick a Metal kernel, but were guarded withmx.is_available(mx.gpu), so they ran on CUDA too. On CUDAGatedDeltaUpdate::use_fallbackalways returns true, so both sides of the comparison were the same unrolled fallback graph over the largegpu_dimscases. That leaves ~12 GiB cached in the allocator on the T4 runner;test_fftthen hits a cuFFT allocation failure, which surfaces ascudaPeekAtLastError() failed: out of memoryand aborts the test process.Pre-fix device footprint of the three tests with the fallback forced on: 12.1 GiB. Post-fix (remaining tests): 0.64 GiB.
I have discovered even another weird thing during this investigation but I will probably raise another PR for it...