Condividi tramite


NlpFixedParameters Class

Definition

Fixed training parameters that won't be swept over during AutoML NLP training.

public class NlpFixedParameters
type NlpFixedParameters = class
Public Class NlpFixedParameters
Inheritance
NlpFixedParameters

Constructors

NlpFixedParameters()

Initializes a new instance of NlpFixedParameters.

Properties

GradientAccumulationSteps

Number of steps to accumulate gradients over before running a backward pass.

LearningRate

The learning rate for the training procedure.

LearningRateScheduler

The type of learning rate schedule to use during the training procedure.

ModelName

The name of the model to train.

NumberOfEpochs

Number of training epochs.

TrainingBatchSize

The batch size for the training procedure.

ValidationBatchSize

The batch size to be used during evaluation.

WarmupRatio

The warmup ratio, used alongside LrSchedulerType.

WeightDecay

The weight decay for the training procedure.

Applies to