CUDA kernel sync fixes, test_layer enhancement #3131
Closed
Conversation
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
…des an optimized linear transformation for multi-dimensional inputs.
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
…tion-free tokenization
…_computation_time layer
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Outdated suggestions cannot be applied.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.
Summary
This PR addresses critical CUDA synchronization issues and enhances the
test_layerutility function.CUDA Kernel Fixes
Several CUDA kernels were using
__syncthreads()for cross-block synchronization, which is incorrect since__syncthreads()only synchronizes threads within the same block, not across different blocks. Whengrid_stride_range_ydistributes work across multiple blocks, these synchronization barriers fail silently.Affected functions decomposed into separate kernels:
inverse_norms()dot_prods()multiply_conv()layer_normalize()rms_normalize()compute_act_halt_probabilities()The fix replaces intra-kernel
__syncthreads()with sequentiallaunch_kernel()calls, which provide implicit synchronization between kernel executions.test_layer Enhancement
Modified
test_layerto accept optional parameters for testing layers with specific tensor input constraints, enabling proper gradient verification for layers that require particular input dimensions.Related Discussion
Follow-up to #3128