Dogacel/auto-gpu-kernel

Winner πŸ† (Agent-only) MLSys 2026 - FlashInfer AI Kernel Generation Contest for the DeepSeek Sparse Attention (DSA) track with an average speedup of 34.93x

This repository also configures its own agents. See what auto-gpu-kernel tells them β†’

316Stars on the repository
11Mods indexed here, across every type
1mo agoLast push, which is what freshness is scored on
Apache-2.0Licence, which decides whether bodies are shown

profiler

01

Dogacel/auto-gpu-kernel

Agent Claude Code

Benchmark the current Triton kernel and identify compute/memory/launch bottlenecks with concrete numbers.

not rated 316 +3 1mo ago A 23 tokens original Apache-2.0

research

02

Dogacel/auto-gpu-kernel

Agent Claude Code

Clean-context diagnosis for planning next experiments.

not rated 316 +3 1mo ago A 9 tokens original Apache-2.0

workload-inspector

03

Dogacel/auto-gpu-kernel

Agent Claude Code

Inspect workload input characteristics (shapes, distributions, padding, locality) and recommend kernel optimizations keyed to them.

not rated 316 +3 1mo ago A 28 tokens original Apache-2.0

workload-inspector

04

Dogacel/auto-gpu-kernel

Agent Claude Code

Characterize the workload inputs β€” shapes, distributions, padding, locality β€” and recommend optimizations keyed to them. Call when you suspect a data-shape angle.

not rated 316 +3 1mo ago A 37 tokens original Apache-2.0

At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page β€” the rest are on their repository pages: