mmpretrain.engine

This package includes some runtime components, including hooks, runners, optimizers and loops. These components are useful in classification tasks but not supported by MMEngine yet.

Note

Some components may be moved to MMEngine in the future.

Hooks

ClassNumCheckHook	Class Number Check HOOK.
PreciseBNHook	Precise BN hook.
VisualizationHook	Classification Visualization Hook.
PrepareProtoBeforeValLoopHook	The hook to prepare the prototype in retrievers.
SetAdaptiveMarginsHook	Set adaptive-margins in ArcFaceClsHead based on the power of category-wise count.
EMAHook	A Hook to apply Exponential Moving Average (EMA) on the model during training.
SimSiamHook	Hook for SimSiam.
DenseCLHook	Hook for DenseCL.
SwAVHook	Hook for SwAV.

Optimizers

Lamb	A pure pytorch variant of FuseLAMB (NvLamb variant) optimizer.
LARS	Implements layer-wise adaptive rate scaling for SGD.
LearningRateDecayOptimWrapperConstructor	Different learning rates are set for different layers of backbone.