AzureMLProgressBar

class health_ml.utils.AzureMLProgressBar(refresh_rate=50, print_timestamp=True, write_to_logging_info=False)[source]

Bases: pytorch_lightning.callbacks.progress.base.ProgressBarBase

A PL progress bar that works better in AzureML. It prints timestamps for each message, and works well with a setup where there is no direct access to the console.

Usage example:

>>> from health_ml.utils import AzureMLProgressBar
>>> from pytorch_lightning import Trainer
>>> progress = AzureMLProgressBar(refresh_rate=100)
>>> trainer = Trainer(callbacks=[progress])

Parameters

refresh_rate (int) – The number of steps after which the progress should be printed out.
print_timestamp (bool) – If True, each message that the progress bar prints will be prefixed with the current time in UTC. If False, no such prefix will be added.
write_to_logging_info (bool) – If True, the progress information will be printed via logging.info. If False, it will be printed to stdout via print.

Attributes Summary

`PROGRESS_STAGE_PREDICT`	A string that indicates that the trainer loop is presently in prediction mode.
`PROGRESS_STAGE_TEST`	A string that indicates that the trainer loop is presently in testing mode.
`PROGRESS_STAGE_TRAIN`	A string that indicates that the trainer loop is presently in training mode.
`PROGRESS_STAGE_VAL`	A string that indicates that the trainer loop is presently in validation mode.
`is_disabled`	rtype `bool`
`is_enabled`	rtype `bool`
`refresh_rate`	rtype `int`

Methods Summary

`disable`()	You should provide a way to disable the progress bar.
`enable`()	You should provide a way to enable the progress bar.
`on_predict_batch_end`(args, *kwargs)	Called when the predict batch ends.
`on_predict_epoch_start`(trainer, pl_module)	Called when the predict epoch begins.
`on_test_batch_end`(args, *kwargs)	Called when the test batch ends.
`on_test_epoch_start`(trainer, pl_module)	Called when the test epoch begins.
`on_train_batch_end`(args, *kwargs)	Called when the train batch ends.
`on_train_epoch_start`(trainer, pl_module)	Called when the train epoch begins.
`on_validation_batch_end`(args, *kwargs)	Called when the validation batch ends.
`on_validation_start`(trainer, pl_module)	Called when the validation loop begins.
`start_stage`(stage, total_num_batches)	Sets the information that a new stage of the PL loop is starting.
`update_progress`(batches_processed)	Writes progress information once the refresh interval is full.

Attributes Documentation

PROGRESS_STAGE_PREDICT = 'Prediction': A string that indicates that the trainer loop is presently in prediction mode.

PROGRESS_STAGE_TEST = 'Testing': A string that indicates that the trainer loop is presently in testing mode.

PROGRESS_STAGE_TRAIN = 'Training': A string that indicates that the trainer loop is presently in training mode.

PROGRESS_STAGE_VAL = 'Validation': A string that indicates that the trainer loop is presently in validation mode.

is_disabled

Return type: bool

is_enabled

Return type: bool

refresh_rate

Return type: int

Methods Documentation

disable()[source]

You should provide a way to disable the progress bar.

The Trainer will call this to disable the output on processes that have a rank different from 0, e.g., in multi-node training.

Return type: None

enable()[source]

You should provide a way to enable the progress bar.

The Trainer will call this in e.g. pre-training routines like the learning rate finder to temporarily enable and disable the main progress bar.

Return type: None

on_predict_batch_end(*args, **kwargs)[source]

Called when the predict batch ends.

Return type: None

on_predict_epoch_start(trainer, pl_module)[source]

Called when the predict epoch begins.

Return type: None

on_test_batch_end(*args, **kwargs)[source]

Called when the test batch ends.

Return type: None

on_test_epoch_start(trainer, pl_module)[source]

Called when the test epoch begins.

Return type: None

on_train_batch_end(*args, **kwargs)[source]

Called when the train batch ends.

Return type: None

on_train_epoch_start(trainer, pl_module)[source]

Called when the train epoch begins.

Return type: None

on_validation_batch_end(*args, **kwargs)[source]

Called when the validation batch ends.

Return type: None

on_validation_start(trainer, pl_module)[source]

Called when the validation loop begins.

Return type: None

start_stage(stage, total_num_batches)[source]

Sets the information that a new stage of the PL loop is starting. The stage will be available in self.stage, total_num_batches in self.total_num_batches. The time when this method was called is recorded in self.stage_start_time

Parameters

stage (str) – The string name of the stage that has just started.
total_num_batches (int) – The total number of batches that need to be processed in this stage. This is used only for progress reporting.

Return type

None

update_progress(batches_processed)[source]

Writes progress information once the refresh interval is full.

Parameters: batches_processed (int) – The number of batches that have been processed for the current stage.
Return type: None