Optimizer Configuration

Configures the optimizer of a deep learning model.

Input and Output Channels

There is one input channel "input" and one output channel "output". During training, the channels are of class ModelDescriptor. During prediction, the channels are of class Task.

State

The state is the value calculated by the public method shapes_out().

Parameters

The parameters are defined dynamically from the optimizer that is set during construction.

Internals

During training, the optimizer is cloned and added to the ModelDescriptor. Note that the parameter set of the stored TorchOptimizer is reference-identical to the parameter set of the pipeop itself.

See also

Other PipeOp: mlr_pipeops_module, mlr_pipeops_torch_callbacks

Other Model Configuration: ModelDescriptor(), mlr_pipeops_torch_callbacks, mlr_pipeops_torch_loss, model_descriptor_union()

Super class

mlr3pipelines::PipeOp -> PipeOpTorchOptimizer

Methods

Public methods

• PipeOpTorchOptimizer$new()

• PipeOpTorchOptimizer$clone()

Inherited methods

• mlr3pipelines::PipeOp$help()
• mlr3pipelines::PipeOp$predict()
• mlr3pipelines::PipeOp$print()
• mlr3pipelines::PipeOp$train()

---

Method new()

Creates a new instance of this R6 class.

Usage

PipeOpTorchOptimizer$new(
  optimizer = t_opt("adam"),
  id = "torch_optimizer",
  param_vals = list()
)

Arguments

optimizer

(TorchOptimizer or character(1) or torch_optimizer_generator)
The optimizer (or something convertible via as_torch_optimizer()).

id

(character(1))
Identifier of the resulting object.

param_vals

(list())
List of hyperparameter settings, overwriting the hyperparameter settings that would otherwise be set during construction.

---

Method clone()

The objects of this class are cloneable with this method.

Usage

PipeOpTorchOptimizer$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

po_opt = po("torch_optimizer", "sgd", lr = 0.01)
po_opt$param_set
#> <ParamSet(5)>
#>              id    class lower upper nlevels        default  value
#>          <char>   <char> <num> <num>   <num>         <list> <list>
#> 1:           lr ParamDbl     0   Inf     Inf <NoDefault[0]>   0.01
#> 2:     momentum ParamDbl     0     1     Inf              0 [NULL]
#> 3:    dampening ParamDbl     0     1     Inf              0 [NULL]
#> 4: weight_decay ParamDbl     0     1     Inf              0 [NULL]
#> 5:     nesterov ParamLgl    NA    NA       2          FALSE [NULL]
mdin = po("torch_ingress_num")$train(list(tsk("iris")))
mdin[[1L]]$optimizer
#> NULL
mdout = po_opt$train(mdin)
mdout[[1L]]$optimizer
#> <TorchOptimizer:sgd> Stochastic Gradient Descent
#> * Generator: optim_sgd
#> * Parameters: lr=0.01
#> * Packages: torch,mlr3torch