Add parallel chunked prefill for channelwise gated delta rule (#21061)#21061
Merged
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21061
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 1b7ba84 with merge base 179c4ee ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
Contributor
|
@JakeStevens has exported this pull request. If you are a Meta employee, you can view the originating Diff in D112597348. |
This PR needs a
|
meta-codesync
Bot
changed the title
JakeStevens
added a commit
to JakeStevens/executorch
that referenced
this pull request
Jul 21, 2026
…h#21061) Summary: Route the channelwise gated delta rule by sequence length: T == 1 keeps the two-pass token recurrence for autoregressive decode, while T != 1 uses a chunkwise WY/UT formulation for prefill. The chunked path computes per-channel log-decay prefixes, causal query/key terms, the beta-folded triangular transform, WY pseudo-keys and pseudo-values, and inter-chunk state carry. It handles a ragged final chunk without a separate tail implementation. Parallelize independent (batch, head) work across the ExecuTorch threadpool. Each worker receives a disjoint slice of one temporary scratch arena, avoiding shared mutable buffers while amortizing allocation across chunks. Reviewed By: billmguo Differential Revision: D112597348
JakeStevens
force-pushed
the
export-D112597348
branch
from
July 21, 2026 14:50
1509cce to
85d15b3
Compare
JakeStevens
added a commit
to JakeStevens/executorch
that referenced
this pull request
Jul 21, 2026
…h#21061) Summary: Route the channelwise gated delta rule by sequence length: T == 1 keeps the two-pass token recurrence for autoregressive decode, while T != 1 uses a chunkwise WY/UT formulation for prefill. The chunked path computes per-channel log-decay prefixes, causal query/key terms, the beta-folded triangular transform, WY pseudo-keys and pseudo-values, and inter-chunk state carry. It handles a ragged final chunk without a separate tail implementation. Parallelize independent (batch, head) work across the ExecuTorch threadpool. Each worker receives a disjoint slice of one temporary scratch arena, avoiding shared mutable buffers while amortizing allocation across chunks. Reviewed By: billmguo Differential Revision: D112597348
JacobSzwejbka
approved these changes
Jul 22, 2026
…BUCK (pytorch#21105) Summary: Relands the two-pass optimization for the `channelwise_gated_delta_rule` custom op (originally pytorch#21020, D112596724), which was reverted in D113048961 because it broke OSS `unittest macos / linux`. The revert was caused by the benchmark BUCK target: ``` runtime.python_binary(name = ..., srcs = [...], main_module = ...) ``` Fix: move the source into a `runtime.python_library` and have the `runtime.python_binary` reference it via `deps` with only `main_module` Differential Revision: D113076546
…h#21061) Summary: Route the channelwise gated delta rule by sequence length: T == 1 keeps the two-pass token recurrence for autoregressive decode, while T != 1 uses a chunkwise WY/UT formulation for prefill. The chunked path computes per-channel log-decay prefixes, causal query/key terms, the beta-folded triangular transform, WY pseudo-keys and pseudo-values, and inter-chunk state carry. It handles a ragged final chunk without a separate tail implementation. Parallelize independent (batch, head) work across the ExecuTorch threadpool. Each worker receives a disjoint slice of one temporary scratch arena, avoiding shared mutable buffers while amortizing allocation across chunks. Reviewed By: billmguo Differential Revision: D112597348
JakeStevens
force-pushed
the
export-D112597348
branch
from
July 22, 2026 19:53
85d15b3 to
abb6ab8
Compare
JakeStevens
added a commit
to JakeStevens/executorch
that referenced
this pull request
Jul 22, 2026
…h#21061) Summary: Route the channelwise gated delta rule by sequence length: T == 1 keeps the two-pass token recurrence for autoregressive decode, while T != 1 uses a chunkwise WY/UT formulation for prefill. The chunked path computes per-channel log-decay prefixes, causal query/key terms, the beta-folded triangular transform, WY pseudo-keys and pseudo-values, and inter-chunk state carry. It handles a ragged final chunk without a separate tail implementation. Parallelize independent (batch, head) work across the ExecuTorch threadpool. Each worker receives a disjoint slice of one temporary scratch arena, avoiding shared mutable buffers while amortizing allocation across chunks. Reviewed By: billmguo Differential Revision: D112597348
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.
Summary:
Route the channelwise gated delta rule by sequence length: T == 1 keeps the two-pass token recurrence for autoregressive decode, while T != 1 uses a chunkwise WY/UT formulation for prefill.
The chunked path computes per-channel log-decay prefixes, causal query/key terms, the beta-folded triangular transform, WY pseudo-keys and pseudo-values, and inter-chunk state carry. It handles a ragged final chunk without a separate tail implementation.
Parallelize independent (batch, head) work across the ExecuTorch threadpool. Each worker receives a disjoint slice of one temporary scratch arena, avoiding shared mutable buffers while amortizing allocation across chunks.
Reviewed By: billmguo
Differential Revision: D112597348