_target_: lightning.pytorch.strategies.FSDPStrategy
sharding_strategy: ${resolve_variable:torch.distributed.fsdp.ShardingStrategy.FULL_SHARD}
cpu_offload: null
activation_checkpointing: null
mixed_precision:
  _target_: torch.distributed.fsdp.MixedPrecision
  param_dtype: null
  reduce_dtype: null
  buffer_dtype: null
  keep_low_precision_grads: false
  cast_forward_inputs: false
  cast_root_forward_inputs: true