Apppublic

aneforge/ane-leaderboard

sourceHugging Facemitupdated 2mo agoView on Hugging Face

0likes

App README

Apple Neural Engine leaderboard

Peak on-engine performance (fp16 GEMM, perf-per-watt, decode) and fp16-correctness limits across Apple Silicon, measured with ANEForge. Data comes from the ane-rooflines dataset and updates as chips are contributed.

Add your Mac (2 minutes, you get credited):

PYTHONPATH=. python3 bench/roofline_suite.py --contributor <your-gh-handle>

then open a PR — see the roofline drive.