Why it matters
ZeRO-1 is first sharding. Understanding shapes ZeRO.
Advertisement
The architecture
Optim state sharded across DP replicas.
Each device owns 1/N optim state.
Grad reduction + optim update on owner.
Advertisement
How it works end to end
Rajbhandari 2020 ZeRO.
Simplest ZeRO stage.
Communication similar to DDP.