Skip to contents

Package website: release | dev

Deep Learning with torch and mlr3.


# Install from CRAN
# Install the development version from GitHub:

Afterwards, you also need to run the command below:


More information about installing torch can be found here.

What is mlr3torch?

mlr3torch is a deep learning framework for the mlr3 ecosystem built on top of torch. It allows to easily build, train and evaluate deep learning models in a few lines of codes, without needing to worry about low-level details. Off-the-shelf learners are readily available, but custom architectures can be defined by connecting PipeOpTorch operators in an mlr3pipelines::Graph.

Using predefined learners such as a simple multi layer perceptron (MLP) works just like any other mlr3 Learner.

learner_mlp = lrn("classif.mlp",
  # defining network parameters
  activation     = nn_relu,
  neurons        = c(20, 20),
  # training parameters
  batch_size     = 16,
  epochs         = 50,
  device         = "cpu",
  # Proportion of data to use for validation
  validate = 0.3,
  # Defining the optimizer, loss, and callbacks
  optimizer      = t_opt("adam", lr = 0.1),
  loss           = t_loss("cross_entropy"),
  callbacks      = t_clbk("history"), # this saves the history in the learner
  # Measures to track
  measures_valid = msrs(c("classif.logloss", "classif.ce")),
  measures_train = msrs(c("classif.acc")),
  # predict type (required by logloss)
  predict_type = "prob"

Below, we train this learner on the sonar example task:


Next, we construct the same architecture using PipeOpTorch objects. The first pipeop – a PipeOpTorchIngress – defines the entrypoint of the network. All subsequent pipeops define the neural network layers.

architecture = po("torch_ingress_num") %>>%
  po("nn_linear", out_features = 20) %>>%
  po("nn_relu") %>>%

To turn this into a learner, we configure the loss, optimizer, callbacks as well as the training arguments.

graph_mlp = architecture %>>%
  po("torch_loss", loss = t_loss("cross_entropy")) %>>%
  po("torch_optimizer", optimizer = t_opt("adam", lr = 0.1)) %>>%
  po("torch_callbacks", callbacks = t_clbk("history")) %>>%
    batch_size = 16, epochs = 50, device = "cpu")

graph_lrn = as_learner(graph_mlp)

To work with generic tensors, the lazy_tensor type can be used. It wraps a torch::dataset, but allows to preprocess the data (lazily) using PipeOp objects. Below, we flatten the MNIST task, so we can then train a multi-layer perceptron on it. Note that this does not transform the data in-memory, but is only applied when the data is actually loaded.

# load the predefined mnist task
mnist = tsk("mnist")
#>     label           image
#>    <fctr>   <lazy_tensor>
#> 1:      5 <tnsr[1x28x28]>
#> 2:      0 <tnsr[1x28x28]>
#> 3:      4 <tnsr[1x28x28]>

# Flatten the images
flattener = po("trafo_reshape", shape = c(-1, 28 * 28))
mnist_flat = flattener$train(list(mnist))[[1L]]

#>     label         image
#>    <fctr> <lazy_tensor>
#> 1:      5   <tnsr[784]>
#> 2:      0   <tnsr[784]>
#> 3:      4   <tnsr[784]>

To actually access the tensors, we can call materialize(). We only show a slice of the resulting tensor for readability:

  mnist_flat$data(1:2, cols = "image")[[1L]],
  rbind = TRUE
)[1:2, 1:4]
#> torch_tensor
#>  0  0  0  0
#>  0  0  0  0
#> [ CPUFloatType{2,4} ]

Below, we define a more complex architecture that has one single input which is a lazy_tensor. For that, we first define a single residual block:

layer = list(
  po("nn_linear", out_features = 50L) %>>%
    po("nn_dropout") %>>% po("nn_relu")
) %>>% po("nn_merge_sum")

Next, we create a neural network that takes as input a lazy_tensor (po("torch_ingress_num")). It first applies a linear layer and then repeats the above layer using the special PipeOpTorchBlock, followed by the network’s head. After that, we configure the loss, optimizer and the training parameters. Note that po("nn_linear_0") is equivalent to po("nn_linear", id = "nn_linear_0") and we need this here to avoid ID clashes with the linear layer from po("nn_block").

deep_network = po("torch_ingress_ltnsr") %>>%
  po("nn_linear_0", out_features = 50L) %>>%
  po("nn_block", layer, n_blocks = 5L) %>>%
  po("nn_head") %>>%
  po("torch_loss", loss = t_loss("cross_entropy")) %>>%
  po("torch_optimizer", optimizer = t_opt("adam")) %>>%
    epochs = 100L, batch_size = 32

Next, we prepend the preprocessing step that flattens the images so we can directly apply this learner to the unflattened MNIST task.

deep_learner = as_learner(
  flattener %>>% deep_network
deep_learner$id = "deep_network"

In order to keep track of the performance during training, we use 20% of the data and evaluate it using classification accuracy.

set_validate(deep_learner, 0.2)
  torch_model_classif.measures_valid = msr("classif.acc")

All that is left is to train the learner:


Feature Overview

  • Off-the-shelf architectures are readily available as mlr3::Learners.
  • Currently, supervised regression and classification is supported.
  • Custom learners can be defined using the Graph language from mlr3pipelines.
  • The package supports tabular data, as well as generic tensors via the lazy_tensor type.
  • Multi-modal data can be handled conveniently, as lazy_tensor objects can be stored alongside tabular data.
  • It is possible to customize the training process via (predefined or custom) callbacks.
  • The package is fully integrated into the mlr3 ecosystem.
  • Neural network architectures, as well as their hyperparameters can be easily tuned via mlr3tuning and friends.


  • Start by reading one of the vignettes on the package website!


  • To run the tests one needs to set the environment variable TEST_TORCH = 1, e.g. by adding it to .Renviron.


  • Without the great R package torch none of this would have been possible.
  • The names for the callback stages are taken from luz, another high-level deep learning framework for R torch.
  • Building neural networks using PipeOpTorch operators is inspired by keras.

Bugs, Questions, Feedback

mlr3torch is a free and open source software project that encourages participation and feedback. If you have any issues, questions, suggestions or feedback, please do not hesitate to open an “issue” about it on the GitHub page!

In case of problems / bugs, it is often helpful if you provide a “minimum working example” that showcases the behaviour (but don’t worry about this if the bug is obvious).

Please understand that the resources of the project are limited: response may sometimes be delayed by a few days, and some feature suggestions may be rejected if they are deemed too tangential to the vision behind the project.