# TODO(yair): Currently not compatible with grad clipping
_target_: lightning.pytorch.strategies.FSDPStrategy
sharding_strategy: SHARD_GRAD_OP