PipeOp Torch Classifier

Builds a torch classifier and trains it.

Parameters

See LearnerTorch

Input and Output Channels

There is one input channel "input" that takes in ModelDescriptor during traing and a Task of the specified task_type during prediction. The output is NULL during training and a Prediction of given task_type during prediction.

State

A trained LearnerTorchModel.

Internals

A LearnerTorchModel is created by calling model_descriptor_to_learner() on the provided ModelDescriptor that is received through the input channel. Then the parameters are set according to the parameters specified in PipeOpTorchModel and its '$train() method is called on the [Task][mlr3::Task] stored in the [ModelDescriptor`].

Other PipeOps: mlr_pipeops_nn_adaptive_avg_pool1d, mlr_pipeops_nn_adaptive_avg_pool2d, mlr_pipeops_nn_adaptive_avg_pool3d, mlr_pipeops_nn_avg_pool1d, mlr_pipeops_nn_avg_pool2d, mlr_pipeops_nn_avg_pool3d, mlr_pipeops_nn_batch_norm1d, mlr_pipeops_nn_batch_norm2d, mlr_pipeops_nn_batch_norm3d, mlr_pipeops_nn_block, mlr_pipeops_nn_celu, mlr_pipeops_nn_conv1d, mlr_pipeops_nn_conv2d, mlr_pipeops_nn_conv3d, mlr_pipeops_nn_conv_transpose1d, mlr_pipeops_nn_conv_transpose2d, mlr_pipeops_nn_conv_transpose3d, mlr_pipeops_nn_dropout, mlr_pipeops_nn_elu, mlr_pipeops_nn_flatten, mlr_pipeops_nn_gelu, mlr_pipeops_nn_glu, mlr_pipeops_nn_hardshrink, mlr_pipeops_nn_hardsigmoid, mlr_pipeops_nn_hardtanh, mlr_pipeops_nn_head, mlr_pipeops_nn_layer_norm, mlr_pipeops_nn_leaky_relu, mlr_pipeops_nn_linear, mlr_pipeops_nn_log_sigmoid, mlr_pipeops_nn_max_pool1d, mlr_pipeops_nn_max_pool2d, mlr_pipeops_nn_max_pool3d, mlr_pipeops_nn_merge, mlr_pipeops_nn_merge_cat, mlr_pipeops_nn_merge_prod, mlr_pipeops_nn_merge_sum, mlr_pipeops_nn_prelu, mlr_pipeops_nn_relu, mlr_pipeops_nn_relu6, mlr_pipeops_nn_reshape, mlr_pipeops_nn_rrelu, mlr_pipeops_nn_selu, mlr_pipeops_nn_sigmoid, mlr_pipeops_nn_softmax, mlr_pipeops_nn_softplus, mlr_pipeops_nn_softshrink, mlr_pipeops_nn_softsign, mlr_pipeops_nn_squeeze, mlr_pipeops_nn_tanh, mlr_pipeops_nn_tanhshrink, mlr_pipeops_nn_threshold, mlr_pipeops_nn_unsqueeze, mlr_pipeops_torch_ingress, mlr_pipeops_torch_ingress_categ, mlr_pipeops_torch_ingress_ltnsr, mlr_pipeops_torch_ingress_num, mlr_pipeops_torch_loss, mlr_pipeops_torch_model, mlr_pipeops_torch_model_regr

Super classes

mlr3pipelines::PipeOp -> mlr3pipelines::PipeOpLearner -> mlr3torch::PipeOpTorchModel -> PipeOpTorchModelClassif

Methods

Inherited methods

Method `new()`

Creates a new instance of this R6 class.

Usage

PipeOpTorchModelClassif$new(id = "torch_model_classif", param_vals = list())

Arguments

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

PipeOpTorchModelClassif$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

# simple logistic regression

# configure the model descriptor
md = as_graph(po("torch_ingress_num") %>>%
  po("nn_head") %>>%
  po("torch_loss", "cross_entropy") %>>%
  po("torch_optimizer", "adam"))$train(tsk("iris"))[[1L]]

print(md)
#> <ModelDescriptor: 2 ops>
#> * Ingress:  torch_ingress_num.input: [(NA,4)]
#> * Task:  iris [classif]
#> * Callbacks:  N/A
#> * Optimizer:  Adaptive Moment Estimation
#> * Loss:  Cross Entropy
#> * pointer:  nn_head.output [(NA,3)]

# build the learner from the model descriptor and train it
po_model = po("torch_model_classif", batch_size = 50, epochs = 1)
po_model$train(list(md))
#> $output
#> NULL
#> 
po_model$state
#> $model
#> $network
#> An `nn_module` containing 15 parameters.
#> 
#> ── Modules ─────────────────────────────────────────────────────────────────────
#> • module_list: <nn_module_list> #15 parameters
#> 
#> $internal_valid_scores
#> NULL
#> 
#> $loss_fn
#> list()
#> 
#> $optimizer
#> $optimizer$param_groups
#> $optimizer$param_groups[[1]]
#> $optimizer$param_groups[[1]]$params
#> [1] 1 2
#> 
#> $optimizer$param_groups[[1]]$lr
#> [1] 0.001
#> 
#> $optimizer$param_groups[[1]]$betas
#> [1] 0.900 0.999
#> 
#> $optimizer$param_groups[[1]]$eps
#> [1] 1e-08
#> 
#> $optimizer$param_groups[[1]]$weight_decay
#> [1] 0
#> 
#> $optimizer$param_groups[[1]]$amsgrad
#> [1] FALSE
#> 
#> 
#> 
#> $optimizer$state
#> $optimizer$state$`1`
#> $optimizer$state$`1`$step
#> [1] 3
#> 
#> $optimizer$state$`1`$exp_avg
#> torch_tensor
#>  0.1824  0.0781  0.0404 -0.0493
#> -0.0506 -0.0145  0.0062  0.0439
#> -0.1318 -0.0637 -0.0466  0.0055
#> [ CPUFloatType{3,4} ]
#> 
#> $optimizer$state$`1`$exp_avg_sq
#> torch_tensor
#> 0.01 *
#>  0.5263  0.0525  2.0401  0.7900
#>   1.1783  0.1205  2.5981  0.6958
#>   1.3648  0.1695  2.2630  0.5236
#> [ CPUFloatType{3,4} ]
#> 
#> 
#> $optimizer$state$`2`
#> $optimizer$state$`2`$step
#> [1] 3
#> 
#> $optimizer$state$`2`$exp_avg
#> torch_tensor
#> 0.001 *
#> -5.6068
#>  3.9102
#>  1.6966
#> [ CPUFloatType{3} ]
#> 
#> $optimizer$state$`2`$exp_avg_sq
#> torch_tensor
#> 0.0001 *
#>  7.1996
#>  7.7785
#>  5.8956
#> [ CPUFloatType{3} ]
#> 
#> 
#> 
#> 
#> $epochs
#> [1] 1
#> 
#> $callbacks
#> named list()
#> 
#> $seed
#> [1] 1285070263
#> 
#> $task_col_info
#> Key: <id>
#>              id    type                      levels
#>          <char>  <char>                      <list>
#> 1: Petal.Length numeric                      [NULL]
#> 2:  Petal.Width numeric                      [NULL]
#> 3: Sepal.Length numeric                      [NULL]
#> 4:  Sepal.Width numeric                      [NULL]
#> 5:      Species  factor setosa,versicolor,virginica
#> 
#> attr(,"class")
#> [1] "learner_torch_model" "list"               
#> 
#> $log
#> Empty data.table (0 rows and 3 cols): stage,class,msg
#> 
#> $train_time
#> [1] 0.055
#> 
#> $param_vals
#> $param_vals$epochs
#> [1] 1
#> 
#> $param_vals$device
#> [1] "auto"
#> 
#> $param_vals$num_threads
#> [1] 1
#> 
#> $param_vals$num_interop_threads
#> [1] 1
#> 
#> $param_vals$seed
#> [1] "random"
#> 
#> $param_vals$eval_freq
#> [1] 1
#> 
#> $param_vals$measures_train
#> list()
#> 
#> $param_vals$measures_valid
#> list()
#> 
#> $param_vals$patience
#> [1] 0
#> 
#> $param_vals$min_delta
#> [1] 0
#> 
#> $param_vals$batch_size
#> [1] 50
#> 
#> 
#> $task_hash
#> [1] "34a6beb27c3181c3"
#> 
#> $feature_names
#> [1] "Petal.Length" "Petal.Width"  "Sepal.Length" "Sepal.Width" 
#> 
#> $validate
#> NULL
#> 
#> $mlr3_version
#> [1] ‘0.21.1’
#> 
#> $internal_tuned_values
#> named list()
#> 
#> $data_prototype
#> Empty data.table (0 rows and 5 cols): Species,Petal.Length,Petal.Width,Sepal.Length,Sepal.Width
#> 
#> $task_prototype
#> Empty data.table (0 rows and 5 cols): Species,Petal.Length,Petal.Width,Sepal.Length,Sepal.Width
#> 
#> $train_task
#> <TaskClassif:iris> (150 x 5): Iris Flowers
#> * Target: Species
#> * Properties: multiclass
#> * Features (4):
#>   - dbl (4): Petal.Length, Petal.Width, Sepal.Length, Sepal.Width
#> 
#> attr(,"class")
#> [1] "learner_state" "list"

Parameters

Input and Output Channels

State

Internals

See also

Super classes

Methods

Public methods

Method new()

Usage

Arguments

Method clone()

Usage

Arguments

Examples

Method `new()`

Method `clone()`