Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.
Add FSDP docs (#7791)
* Add FSDP docs * Address reviews * Add note about how FSDP can replace pipe parallelism * Add import * Remove sentence
S
Sean Naren committed
0a72fd2284eb9e09afefc74886187e20935fbfa4
Parent: e4ba06c
Committed by GitHub <[email protected]>
on 6/2/2021, 9:52:48 AM