Deep.Net


Train

Namespace: Models

Generic training module.

Nested types and modules

TypeDescription
Cfg

Training configuration.

Faith

training faith

ITrainable<'Smpl, 'T>

Interface for a trainable model.

TerminationCriterium

Training termination criterium

TrainingResult

Result of training

Functions and values

Function or valueDescription
defaultCfg
Signature: Cfg

Default training configuration.

newStatefulTrainable (...)
Signature: modelInstance:ModelInstance<'?175955> -> losses:ExprT list -> nextState:ExprT -> varEnvBuilder:(Tensor<'?175955> option -> '?175956 -> VarEnvT) -> optNew:(ExprT -> ExprT -> IDevice -> IOptimizer<'?175955,'?175957,'?175958>) -> optCfg:'?175957 -> ITrainable<'?175956,'?175955>
Type parameters: '?175955, '?175956, '?175957, '?175958

Constructs an ITrainable<_> for the given stateful model using the specified loss expressions, state update expression and optimizer.

train trainable dataset cfg
Signature: trainable:ITrainable<'Smpl,'T> -> dataset:TrnValTst<'Smpl> -> cfg:Cfg -> TrainingResult
Type parameters: 'Smpl, 'T

Trains a model instance using the given loss and optimization functions on the given dataset. Returns the training history.

trainableFromLossExpr (...)
Signature: modelInstance:ModelInstance<'?175965> -> loss:ExprT -> varEnvBuilder:('?175966 -> VarEnvT) -> optNew:(ExprT -> ExprT -> IDevice -> IOptimizer<'?175965,'?175967,'?175968>) -> optCfg:'?175967 -> ITrainable<'?175966,'?175965>
Type parameters: '?175965, '?175966, '?175967, '?175968

Constructs an ITrainable<_> for the given model instance, loss expression and optimizer.

trainableFromLossExprs (...)
Signature: modelInstance:ModelInstance<'?175960> -> losses:ExprT list -> varEnvBuilder:('?175961 -> VarEnvT) -> optNew:(ExprT -> ExprT -> IDevice -> IOptimizer<'?175960,'?175962,'?175963>) -> optCfg:'?175962 -> ITrainable<'?175961,'?175960>
Type parameters: '?175960, '?175961, '?175962, '?175963

Constructs an ITrainable<_> for the given model instance, loss expressions and optimizer.

Fork me on GitHub