Model techniques map
Techniquesoptimizationtraining parallelism

specific method · filed under optimization

ZeRO-3 Optimizer State Sharding

A ZeRO-3 method that shards optimizer state across ranks.

source
1
model
1
lab adopt it
1
strongest
used

How sources treat it

One count per evidence span, weakest treatment to strongest.

used 1

Documented in

Evidence

1 span quoted from the sources, strongest treatment first.

The optimizer state is sharded using an implementation of ZeRO-3 [Ren et al., 2021].

usedsoftware implementationin Gemma 4Google DeepMind

Filed alongside

Other methods under optimization :: training parallelism.