library("rplatypus")
callback_early_stopping(monitor = "val_dice", patience = 10)$name
[1] "early_stopping"
$monitor
[1] "val_dice"
$patience
[1] 10
$min_delta
[1] 0
$restore_best
[1] TRUECallbacks
Description
Things that happen between epochs. monitor is checked against the metrics the model actually asks for, so a callback watching a number that will never arrive is refused when the specification is built rather than waited on forever.
Usage
callback_early_stopping(
monitor = "val_loss",
patience = 10,
min_delta = 0,
restore_best = TRUE
)
callback_model_checkpoint(path, monitor = "val_loss", save_best_only = TRUE)
callback_reduce_lr_on_plateau(
monitor = "val_loss",
factor = 0.1,
patience = 5,
min_lr = 0
)
callback_cosine_annealing(min_lr = 0, epochs = NULL)
callback_csv_logger(path)
callback_terminate_on_nan()
Arguments
monitor
|
What to watch: “val_loss”, “train_loss”, or “val_” followed by a metric name, such as “val_dice”.
|
patience
|
Epochs to wait before acting. |
min_delta
|
Improvement smaller than this does not count. |
restore_best
|
Put the best weights back when training stops, rather than keeping the last - which are by definition the ones that were not improving. On by default. Note that callback_swa() replaces nothing if a run stops before averaging began.
|
path
|
Where to write. For callback_checkpoint() that is the weights file; for callback_csv_logger() it is one row per epoch appended as the epoch ends - the same numbers training_history() returns, so a run that is interrupted still leaves its history behind. Re-read against the engine’s own text when it changed in 0.8.0a23, which is what the field-prose ledger is for: the engine’s sentence had named this package’s training_history() to a Python reader, and the fix there left R’s @param saying three words about two different files.
|
save_best_only
|
Only write when the watched quantity improves, which is the default. FALSE writes every epoch, which is what to use when a run may be interrupted rather than when the best epoch is what you want to keep.
|
factor, min_lr
|
For the learning rate schedule. min_lr is the floor for both callback_reduce_lr_on_plateau() and callback_cosine_annealing().
|
epochs
|
For callback_cosine_annealing(), how many epochs to spread the decay over. Unset means the model’s epochs.
|
Details
Whether the watched quantity should rise or fall is worked out from its name: anything ending in loss is minimised, everything else maximised.
There is a seventh, documented separately because it needs a page of its own: callback_swa() averages the weights over the last part of the run. It watches nothing and changes how reproducible a run is rather than how well it scores.
Value
A callback specification.
Cosine annealing
callback_cosine_annealing() decays the learning rate from its initial value to min_lr along a cosine. It watches nothing: it is a function of how far through the run you are rather than of how the run is going, which is the difference from callback_reduce_lr_on_plateau() - and a run may legitimately want both.
Each parameter group decays from its own initial rate, so an encoder_learning_rate set on a model is not flattened at the first epoch. That would have been invisible: the history records one learning_rate, the first group’s.
epochs unset means the model’s own epochs, which is what you want - a cosine that ends where the run ends.