RecomputeStrategy

Enums · AiDotNet.DistributedTraining

Strategy for recomputing activations during the backward pass.

Fields

Field Summary
RecomputeStrategy Full Recompute all activations between the two nearest checkpoints during backward.
RecomputeStrategy None No recomputation.
RecomputeStrategy Selective Only recompute activations that are needed for the current backward step.
int value__