Bump batched MoE BLOCK_M from 16 to 64 on top of persistent kernel#19221
Draft
Gasoonjia wants to merge 1 commit intopersistent-on-hoistfrom
Draft
Bump batched MoE BLOCK_M from 16 to 64 on top of persistent kernel#19221Gasoonjia wants to merge 1 commit intopersistent-on-hoistfrom
Gasoonjia wants to merge 1 commit intopersistent-on-hoistfrom
Conversation
Microbenchmarked on Qwen3.5 MoE prefill (M=1696, top_k=8, 256 experts):
BLOCK_M=16: 3.62 ms
BLOCK_M=32: 2.85 ms (1.27x)
BLOCK_M=64: 2.75 ms (1.32x)
E2E (Qwen3.5-35B-A3B prefill, --moe-activation-dtype int8 --dense-prefill
dequant --cuda_graph, p=1600 d=512, run_1..5 median):
BLOCK_M=16: 5897 tok/s prefill (273 ms), 98.1 tok/s decode
BLOCK_M=64: 6793 tok/s prefill (237 ms), 98.1 tok/s decode
Speedup: 1.152x prefill, decode unchanged (decode uses non-batched
fused_moe kernel)
Outputs are bit-identical between BLOCK_M=16 and BLOCK_M=64 in the
microbenchmark (max abs diff = 0).
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/19221
Note: Links to docs will display an error until the docs builds have been completed. ❌ 8 New Failures, 6 Cancelled JobsAs of commit 015476d with merge base cb4e5ae ( NEW FAILURES - The following jobs have failed:
CANCELLED JOBS - The following jobs were cancelled. Please retry:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Microbenchmarked on Qwen3.5 MoE prefill (M=1696, top_k=8, 256 experts):
BLOCK_M=16: 3.62 ms
BLOCK_M=32: 2.85 ms (1.27x)
BLOCK_M=64: 2.75 ms (1.32x)
E2E (Qwen3.5-35B-A3B prefill, --moe-activation-dtype int8 --dense-prefill dequant --cuda_graph, p=1600 d=512, run_1..5 median):
BLOCK_M=16: 5897 tok/s prefill (273 ms), 98.1 tok/s decode
BLOCK_M=64: 6793 tok/s prefill (237 ms), 98.1 tok/s decode
Speedup: 1.152x prefill, decode unchanged (decode uses non-batched
fused_moe kernel)
Outputs are bit-identical between BLOCK_M=16 and BLOCK_M=64 in the microbenchmark (max abs diff = 0).