Skip to content

Speed up when having padding tokens in DeepEP #6175

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 28 commits into from
May 17, 2025
Merged

Speed up when having padding tokens in DeepEP #6175

merged 28 commits into from
May 17, 2025

Conversation

fzyzcjy
Copy link
Collaborator

@fzyzcjy fzyzcjy commented May 10, 2025

Motivation

test

PYTHONUNBUFFERED=1 SGLANG_TORCH_PROFILER_DIR=/host_home/temp_sglang_server2local python3 -m sglang.launch_server --model-path /dev/shm/DeepSeek-R1 --trust-remote-code --dist-init-addr 192.168.0.55:5757 --nnodes 2 --node-rank ${MY_NODE_RANK} --tp-size ${num_gpu} --dp-size ${num_gpu} --enable-dp-attention --mem-fraction-static 0.8 --chunked-prefill-size $((128*${num_gpu})) --max-running-requests $((${num_gpu}*128)) --context-length 4096 --disable-radix-cache --enable-deepep-moe --deepep-mode low_latency --cuda-graph-bs 128 --decode-log-interval 1

python3 -m sglang.bench_one_batch_server --model-path /dev/shm/DeepSeek-R1 --base-url http://localhost:30000 --batch-size 16 --input-len 1 --output-len 2048 --skip-warmup
  • baseline: 6 tok/s/gpu
  • PR: 29 tok/s/gpu

Modifications

Checklist

ch-wan
ch-wan reviewed May 10, 2025
View reviewed changes
@fzyzcjy fzyzcjy marked this pull request as draft May 12, 2025 00:04
@fzyzcjy fzyzcjy force-pushed the feat/padding_moe branch from 8797942 to 3fecc76 Compare May 12, 2025 00:09
@fzyzcjy fzyzcjy marked this pull request as ready for review May 12, 2025 00:09
@zhyncs zhyncs merged commit 2716830 into sgl-project:main May 17, 2025
113 of 128 checks passed
@lambert0312
Copy link
Contributor

This pr will significantly reduce DeepSeek's inference performance (15%+). Need to look at the specific reasons.

@fzyzcjy
Copy link
Collaborator Author

fzyzcjy commented May 20, 2025

@lambert0312 Looks bad. Could you please show your commands, and would be great to have a profile. My first guess is that, we need to fuse it.

@lambert0312 lambert0312 mentioned this pull request May 21, 2025
Merged
6 tasks
@lambert0312
Copy link
Contributor

@lambert0312 Looks bad. Could you please show your commands, and would be great to have a profile. My first guess is that, we need to fuse it.

@fzyzcjy I tried to modify it. You can see the PR I linked above. Thank you.

@fzyzcjy
Copy link
Collaborator Author

fzyzcjy commented May 21, 2025

Interesting, I thought this line already makes no extra kernels are executed.

image

@fzyzcjy fzyzcjy mentioned this pull request May 27, 2025
Merged
6 tasks
Layssy pushed a commit to Layssy/sglang-iaas that referenced this pull request Jun 9, 2025
@fzyzcjy @Layssy
Speed up when having padding tokens in DeepEP (sgl-project#6175)
26de0da
xwu-intel pushed a commit to xwu-intel/sglang that referenced this pull request Jun 17, 2025
@fzyzcjy @xwu-intel
Speed up when having padding tokens in DeepEP (sgl-project#6175)
376abc1
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ch-wan ch-wan ch-wan approved these changes

@merrymercy merrymercy Awaiting requested review from merrymercy merrymercy is a code owner

@Ying1123 Ying1123 Awaiting requested review from Ying1123 Ying1123 is a code owner

@hnyls2002 hnyls2002 Awaiting requested review from hnyls2002 hnyls2002 is a code owner

@zhyncs zhyncs Awaiting requested review from zhyncs zhyncs is a code owner

@ispobock ispobock Awaiting requested review from ispobock ispobock is a code owner

@ByronHsu ByronHsu Awaiting requested review from ByronHsu

@HaiShaw HaiShaw Awaiting requested review from HaiShaw HaiShaw is a code owner

Assignees
No one assigned
Labels
high priority
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@fzyzcjy @lambert0312 @ch-wan @zhyncs