Skip to content

Optimize triton attention custom mask #3731

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 2 commits into from
Feb 20, 2025
Merged

Optimize triton attention custom mask #3731

merged 2 commits into from
Feb 20, 2025

Conversation

ispobock
Copy link
Collaborator

@ispobock ispobock commented Feb 20, 2025 •
edited
Loading

Motivation

Skip custom mask for prefix part of triton attention to accelerate target verify stage.

python3 -m sglang.launch_server --model meta-llama/Llama-2-7b-chat-hf  --speculative-algo EAGLE --speculative-draft lmzheng/sglang-EAGLE-llama2-chat-7B --speculative-num-steps 2 --speculative-eagle-topk 4 --speculative-num-draft-tokens 4 --disable-radix --attention-backend triton
python3 benchmark/gsm8k/bench_sglang.py --num-questions 1319 --parallel 1319

# main
Accuracy: 0.233
Invalid: 0.002
Latency: 187.807 s
Output throughput: 796.177 token/s

# this pr
Accuracy: 0.233
Invalid: 0.002
Latency: 105.266 s
Output throughput: 1421.813 token/s

@zhyncs zhyncs merged commit ddcf9fe into sgl-project:main Feb 20, 2025
16 of 19 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

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

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

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

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

Successfully merging this pull request may close these issues.
2 participants
@ispobock @zhyncs