Evaluates a fitted model on a datasetSource:
Evaluates a fitted model on a dataset
evaluate( object, data, ..., callbacks = list(), accelerator = NULL, verbose = NULL, dataloader_options = NULL )
A fitted model to evaluate.
(dataloader, dataset or list) A dataloader created with
torch::dataloader()used for training the model, or a dataset created with
torch::dataset()or a list. Dataloaders and datasets must return a list with at most 2 items. The first item will be used as input for the module and the second will be used as a target for the loss function.
(list, optional) A list of callbacks defined with
luz_callback()that will be called during the training procedure. The callbacks
luz_callback_train_valid()are always added by default.
(accelerator, optional) An optional
accelerator()object used to configure device placement of the components like nn_modules, optimizers and batches of data.
(logical, optional) An optional boolean value indicating if the fitting procedure should emit output to the console during training. By default, it will produce output if
TRUE, otherwise it won't print to the console.
Options used when creating a dataloader. See
shuffle=TRUEby default for the training data and
batch_size=32by default. It will error if not
datais already a dataloader.
Once a model has been trained you might want to evaluate its performance
on a different dataset. For that reason, luz provides the
function that takes a fitted model and a dataset and computes the
metrics attached to the model.
Evaluate returns a
luz_module_evaluation object that you can query for
metrics using the
get_metrics function or simply
evaluation <- fitted %>% evaluate(data = valid_dl) metrics <- get_metrics(evaluation) print(evaluation)