TabFM Auto — AutoTabFM • mlr3automl

Tabfm auto.

Tabfm predicts in context, so every prediction runs the backbone over the training rows once per estimator. This is too slow to be useful on the CPU, so the auto is registered for "cuda" only and Auto$check() removes it from the search space whenever devices does not include "cuda". Construct it with AutoTabFM$new(devices = c("cpu", "cuda")) and re-register it in mlr_auto to run it on the CPU anyway.

Value

Object of class R6::R6Class and AutoTabFM.

Python learners

Python learners like TabPFN, TabFM, and fastai run via reticulate and therefore need a Python installation with their required packages. There are two ways to provide it:

Do nothing and let reticulate::py_require() install the required packages into an ephemeral virtual environment automatically.
Point the RETICULATE_PYTHON environment variable to a Python installation that has the required packages installed.

We recommend option 2 when running on many workers, as it avoids the overhead of downloading and installing the packages on each worker. Use install_python_learners() to create a conda environment with the required packages and set RETICULATE_PYTHON to the returned Python binary.

The TabPFN learner additionally requires the TABPFN_TOKEN environment variable to download the model weights.

Super class

Auto -> AutoTabFM

Methods

Inherited methods

`AutoTabFM$new()`

Creates a new instance of this R6 class.

Usage

AutoTabFM$new(id = "tabfm", devices = "cuda")

Arguments

id: (character(1))
Identifier for the new instance.
devices: (character())
Devices the auto is allowed to run on. Defaults to "cuda" only, because tabfm is too slow to be useful on the CPU.

`AutoTabFM$check()`

Check if the auto is compatible with the task.

Usage

AutoTabFM$check(
  task,
  memory_limit = Inf,
  large_data_set = FALSE,
  devices = "cpu"
)

Arguments

task: (mlr3::Task).
memory_limit: (integer(1)).
large_data_set: (logical(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoTabFM$graph()`

Create the graph for the auto.

Usage

AutoTabFM$graph(
  task,
  measure,
  n_threads,
  timeout,
  devices,
  isolate_python = TRUE
)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoTabFM$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoTabFM$estimate_memory(task, devices = "cpu")

Arguments

task: (mlr3::Task).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoTabFM$design_default()`

Default hyperparameters for the learner.

Usage

AutoTabFM$design_default(task)

Arguments

task: (mlr3::Task).

`AutoTabFM$search_space()`

Get the search space for the auto.

Usage

AutoTabFM$search_space(task)

Arguments

task: (mlr3::Task).

`AutoTabFM$clone()`

The objects of this class are cloneable with this method.

Usage

AutoTabFM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("tabfm")
#> <AutoTabFM>
#>   Inherits from: <Auto>
#>   Public:
#>     check: function (task, memory_limit = Inf, large_data_set = FALSE, devices = "cpu") 
#>     clone: function (deep = FALSE) 
#>     design_default: function (task) 
#>     design_set: function (task, measure, size) 
#>     devices: cuda
#>     early_stopping_rounds: function (task, budget = Inf) 
#>     estimate_memory: function (task, devices = "cpu") 
#>     finalize_model: function (graph_learner) 
#>     graph: function (task, measure, n_threads, timeout, devices, isolate_python = TRUE) 
#>     id: tabfm
#>     initialize: function (id = "tabfm", devices = "cuda") 
#>     n_cpu: 1
#>     n_gpu: 1
#>     packages: mlr3 mlr3extralearners callr
#>     properties: 
#>     search_space: function (task) 
#>     task_types: classif regr
#>   Private:
#>     .default_values: list
#>     .search_space: ParamSet, R6