Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.
Add new args to `BatchSizeFinder` (#21163)
* new args to batch size scaler * add to tuner * add to callback * add testing * update * fix tests * Apply suggestion from @deependujha * Apply suggestion from @deependujha * Apply suggestion from @deependujha * Apply suggestion from @deependujha * update * safe default * update * add assertion * fix doc issue --------- Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com> Co-authored-by: Deependu <deependujha21@gmail.com>
N
Nicki Skafte Detlefsen committed
bd27dbf68944f09a4ddf5144f48453c870c5d2ef
Parent: a7cb33a
Committed by GitHub <noreply@github.com>
on 11/6/2025, 10:22:08 AM