Torch Regression Model

Builds a torch regression model 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_ft_cls, mlr_pipeops_nn_ft_transformer_block, mlr_pipeops_nn_geglu, 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_identity, 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_reglu, 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_tokenizer_categ, mlr_pipeops_nn_tokenizer_num, 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_classif

Super classes

mlr3pipelines::PipeOp -> mlr3pipelines::PipeOpLearner -> PipeOpTorchModel -> PipeOpTorchModelRegr

Methods

Inherited methods

`PipeOpTorchModelRegr$new()`

Creates a new instance of this R6 class.

Usage

PipeOpTorchModelRegr$new(id = "torch_model_regr", 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.

`PipeOpTorchModelRegr$clone()`

The objects of this class are cloneable with this method.

Usage

PipeOpTorchModelRegr$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

# simple linear regression

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

print(md)
#> <ModelDescriptor: 2 ops>
#> * Ingress:  torch_ingress_num.input: [(NA,10)]
#> * Task:  mtcars [regr]
#> * Callbacks:  N/A
#> * Optimizer:  Adaptive Moment Estimation
#> * Loss:  Mean Squared Error
#> * pointer:  nn_head.output [(NA,1)]

# build the learner from the model descriptor and train it
po_model = po("torch_model_regr", batch_size = 20, epochs = 1)
po_model$train(list(md))
#> $output
#> NULL
#> 
po_model$state
#> $model
#> $network
#> An `nn_module` containing 11 parameters.
#> 
#> ── Modules ─────────────────────────────────────────────────────────────────────
#> • module_list: <nn_module_list> #11 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]]$weight_decay
#> [1] 0
#> 
#> $optimizer$param_groups[[1]]$betas
#> [1] 0.900 0.999
#> 
#> $optimizer$param_groups[[1]]$eps
#> [1] 1e-08
#> 
#> $optimizer$param_groups[[1]]$amsgrad
#> [1] FALSE
#> 
#> 
#> 
#> $optimizer$state
#> $optimizer$state$`1`
#> $optimizer$state$`1`$exp_avg
#> torch_tensor
#> Columns 1 to 7    3.8876    55.9380   119.2470  5099.1304    54.3390    56.4534  3133.1069
#> 
#> Columns 8 to 10  273.3301     2.5841    60.2493
#> [ CPUFloatType{1,10} ]
#> 
#> $optimizer$state$`1`$exp_avg_sq
#> torch_tensor
#> Columns 1 to 6 8.3462e-01  1.8993e+02  8.1648e+02  1.5053e+06  1.6550e+02  1.7935e+02
#> 
#> Columns 7 to 10 5.6990e+05  4.2362e+03  4.3138e-01  2.1457e+02
#> [ CPUFloatType{1,10} ]
#> 
#> $optimizer$state$`1`$max_exp_avg_sq
#> torch_tensor
#> [ CPUFloatType{0} ]
#> 
#> $optimizer$state$`1`$step
#> torch_tensor
#>  2
#> [ CPULongType{1} ]
#> 
#> 
#> $optimizer$state$`2`
#> $optimizer$state$`2`$exp_avg
#> torch_tensor
#>  16.0822
#> [ CPUFloatType{1} ]
#> 
#> $optimizer$state$`2`$exp_avg_sq
#> torch_tensor
#>  14.6744
#> [ CPUFloatType{1} ]
#> 
#> $optimizer$state$`2`$max_exp_avg_sq
#> torch_tensor
#> [ CPUFloatType{0} ]
#> 
#> $optimizer$state$`2`$step
#> torch_tensor
#>  2
#> [ CPULongType{1} ]
#> 
#> 
#> 
#> 
#> $epochs
#> [1] 1
#> 
#> $callbacks
#> named list()
#> 
#> $seed
#> [1] 568237855
#> 
#> $task_col_info
#>         id    type levels
#>     <char>  <char> <list>
#>  1:     am numeric [NULL]
#>  2:   carb numeric [NULL]
#>  3:    cyl numeric [NULL]
#>  4:   disp numeric [NULL]
#>  5:   drat numeric [NULL]
#>  6:   gear numeric [NULL]
#>  7:     hp numeric [NULL]
#>  8:   qsec numeric [NULL]
#>  9:     vs numeric [NULL]
#> 10:     wt numeric [NULL]
#> 11:    mpg numeric [NULL]
#> 
#> attr(,"class")
#> [1] "learner_torch_model" "list"               
#> 
#> $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] 20
#> 
#> $param_vals$shuffle
#> [1] TRUE
#> 
#> $param_vals$tensor_dataset
#> [1] FALSE
#> 
#> $param_vals$jit_trace
#> [1] FALSE
#> 
#> 
#> $log
#> Empty data.table (0 rows and 3 cols): stage,class,condition
#> 
#> $train_time
#> elapsed 
#>   0.048 
#> 
#> $task_hash
#> [1] "c7c4f02878d51895"
#> 
#> $feature_names
#>  [1] "am"   "carb" "cyl"  "disp" "drat" "gear" "hp"   "qsec" "vs"   "wt"  
#> 
#> $validate
#> NULL
#> 
#> $mlr3_version
#> [1] ‘1.7.1’
#> 
#> $internal_tuned_values
#> named list()
#> 
#> $data_prototype
#> Empty data.table (0 rows and 11 cols): mpg,am,carb,cyl,disp,drat...
#> 
#> $train_task
#> 
#> ── <TaskRegr> (32x11): Motor Trends ────────────────────────────────────────────
#> • Target: mpg
#> • Properties: -
#> • Features (10):
#>   • dbl (10): am, carb, cyl, disp, drat, gear, hp, qsec, vs, wt
#> 
#> attr(,"class")
#> [1] "learner_state" "list"

Parameters

Input and Output Channels

State

Internals

See also

Super classes

Methods

Public methods

PipeOpTorchModelRegr$new()

Usage

Arguments

PipeOpTorchModelRegr$clone()

Usage

Arguments

Examples

`PipeOpTorchModelRegr$new()`

`PipeOpTorchModelRegr$clone()`