Liger-Kernel icon indicating copy to clipboard operation
Liger-Kernel copied to clipboard

Published 20 hours ago • verified publisher icon linkedin

Choice of num_warps

Open Edenzzzz opened this issue 1 year ago • 5 comments
trafficstars

I wonder why you use 32 (256 threads/block instance) here instead of deciding based on hidden size? Thanks. https://github.com/linkedin/Liger-Kernel/blob/dd86cbd2092177681acf75643ded1b23a785a816/src/liger_kernel/ops/fused_linear_cross_entropy.py#L95

Edenzzzz avatar Sep 22 '24 03:09 Edenzzzz