[ExecuTorch][WebGPU] et_vk.apply_rotary_emb_hf: HuggingFace rotate-half RoPE runtime op (unblocks Qwen3)#20875
Open
JCNTH wants to merge 3 commits into
Open
[ExecuTorch][WebGPU] et_vk.apply_rotary_emb_hf: HuggingFace rotate-half RoPE runtime op (unblocks Qwen3)#20875JCNTH wants to merge 3 commits into
JCNTH wants to merge 3 commits into
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20875
Note: Links to docs will display an error until the docs builds have been completed. ❌ 129 New Failures, 3 Unrelated Failures, 7 Unclassified FailuresAs of commit 3a8fa4b with merge base c2b273e ( NEW FAILURES - The following jobs have failed:
UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:
FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
BROKEN TRUNK - The following job 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. |
This was referenced Jul 10, 2026
[ExecuTorch][WebGPU] Add relu op (shared unary handler; sigmoid adopts make_compute_pipeline)
#20863
Open
This was referenced Jul 10, 2026
This PR needs a
|
meta-cla
Bot
added
the
CLA Signed
This was referenced Jul 16, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Stack from ghstack (oldest at bottom):
Problem: the WebGPU runtime registers only
et_vk.apply_rotary_emb(the interleaved/Meta RoPE convention). HuggingFace-derived models (Qwen3, etc.) export the rotate-half convention, which fuses under VulkanPartitioner intoet_vk.apply_rotary_emb_hf— an op the runtime graph builder has no handler for, soWebGPUGraph::build()throws and the delegate is rejected at load withDelegateInvalidCompatibility(et_load error 48). The whole model then fails to load on WebGPU.Solution: add the
et_vk.apply_rotary_emb_hfruntime kernel + handler as a rotate-half sibling of the interleaved op.Before: only
apply_rotary_emb(interleaved) is registered; HF-RoPE models throw at load.After: both conventions are handled; HF-RoPE models (Qwen3) load and run.
Implementation:
rotary_embedding_hf.wgsl: one thread per (i, i+half_dim) pair (rotate-half pairing vs the interleaved even/odd), reading a full[max_seq, rotary_dim]freqs table indexed at rowstart_pos + s. Scalar,wg_size64 — structural + optimization parity with the interleaved kernel (RoPE is ~1% of runtime; vec4 is neutral for this elementwise-class op on Apple's scalar ALU).RotaryEmbedding.cpp:apply_rotary_emb_hf_implmirrors the interleaved handler; it parses the extrastart_posarg as a build-time Int (baked) or a runtime SymInt (dynamic KV-cache decode) exactly asSdpa.cpphandlesinput_pos, and registers a seq resize hook (xq/xk) plus a start_pos resize hook (dynamic decode). Full rotary only (rotary_dim == head_dim); partial-rotary passthrough throws (documented follow-up; Qwen3 uses full RoPE). Mirrors Vulkanet_vk.apply_rotary_emb_hf(backends/vulkan/runtime/graph/ops/impl/RotaryEmbedding.cpp).et_vk.apply_rotary_emb_hf.default.Constraints: full rotary only for now; scalar one-thread-per-pair, kept at parity with the interleaved sibling rather than vec4 (neutral for RoPE per the closed vec4 sweep).
Co-authored-with: Claude Code.
@exported-using-ghexport
Differential Revision: D111009173
Differential Revision: D111009173