Callback Config

When True, the checkpoints filenames will contain the metric name. Note that val_loss is already embedded in the default filename pattern and enabling this field will produce redundant metric names in the filename.

Number of epochs between checkpoints.

Number of training steps between checkpoints.

One of {min, max}. If save_top_k != 0, the decision to overwrite the current save file is made based on either the maximization or the minimization of the monitored quantity. For 'val_acc', this should be 'max', for 'val_loss' this should be 'min', etc.

Quantity to monitor, currently only val_loss.

When True, saves a {experiment_name}_last.ckpt copy whenever a checkpoint file gets saved.

If save_top_k == k, the best k models according to the quantity monitored will be saved. Ifsave_top_k == 0, no models are saved. ifsave_top_k == -1`, all models are saved.

When True, only the model's weights will be saved (model.save_weights).

Checkpoints are monitored at the specified time interval.

Verbosity mode.

When True, stops training when the monitored quantity becomes NaN or inf.

Whether to run early stopping at the end of the training epoch. If this is False, then the check runs at the end of the validation.

Stop training as soon as the monitored quantity becomes worse than this threshold.

When set True, logs the status of the early stopping callback only for rank 0 process.

Minimum change in the monitored quantity to qualify as an improvement, i.e. an absolute change of less than or equal to min_delta, will count as no improvement.

One of {min, max, auto}.

Quantity to monitor.

Number of checks with no improvement after which training will be stopped.

Stop training immediately once the monitored quantity reaches this threshold.

Verbosity mode.