Skip to main content

Training Configuration Parameters

Audio Classification & Gesture Recognition

ParameterDescription
Number of EpochsTotal training iterations over the dataset
PatienceEpochs to wait before early stopping if validation loss does not improve
Learning RateStep size for weight updates during optimisation
Enable QATEnables Quantization Aware Training
Activation BitsBit width for activation quantisation
Weight BitsBit width for weight quantisation

Image Classification

ParameterDescriptionDefault
EpochsTotal complete passes through the dataset100
PatienceEpochs without improvement before auto-stop4
Batch SizeImages per training step — reduce if out of memory16