noether.core.callbacks.early_stoppers¶

Submodules¶

Exceptions¶

EarlyStopIteration

Custom StopIteration exception for Early Stoppers.

Classes¶

`EarlyStopperBase`	Base class for early stoppers that is used to define the interface for early stoppers used by the trainers.
`FixedEarlyStopper`	Early stopper (training) based on a fixed number of epochs, updates, or samples.
`FixedEarlyStopperConfig`
`MetricEarlyStopper`	Early stopper (training) based on a metric value to be monitored.
`MetricEarlyStopperConfig`	Internal base class for all registry-based configs.

Package Contents¶

exception noether.core.callbacks.early_stoppers.EarlyStopIteration¶

Bases: StopIteration

Custom StopIteration exception for Early Stoppers.

Initialize self. See help(type(self)) for accurate signature.

class noether.core.callbacks.early_stoppers.EarlyStopperBase(callback_config, trainer, model, data_container, tracker, log_writer, checkpoint_writer, metric_property_provider, name=None)¶

Bases: noether.core.callbacks.periodic.PeriodicCallback

Base class for early stoppers that is used to define the interface for early stoppers used by the trainers.

Parameters:

callback_config (noether.core.callbacks.base.CallBackBaseConfig) – Configuration for the callback. See CallBackBaseConfig for available options.
trainer – Trainer of the current run.
model (noether.core.models.ModelBase) – Model of the current run.
data_container (noether.data.container.DataContainer) – DataContainer instance that provides access to all datasets.
tracker (noether.core.trackers.BaseTracker) – BaseTracker instance to log metrics to stdout/disk/online platform.
log_writer (noether.core.writers.LogWriter) – LogWriter instance to log metrics.
checkpoint_writer (noether.core.writers.CheckpointWriter) – CheckpointWriter instance to save checkpoints.
metric_property_provider (noether.core.providers.MetricPropertyProvider) – MetricPropertyProvider instance to access properties of metrics.
name (str | None) – Name of the callback.

to_short_interval_string()¶

Convert the interval to a short string representation used for logging.

Return type:: str

periodic_callback(*, interval_type, update_counter, **kwargs)¶

Check if training should stop and raise exception if needed.

Parameters:

interval_type (noether.core.callbacks.periodic.IntervalType) – Type of interval that triggered this callback.
update_counter (noether.core.utils.training.UpdateCounter) – UpdateCounter instance with current training state.
**kwargs – Additional keyword arguments.

Raises:

EarlyStopIteration – If training should be stopped based on the stopping criterion.

Return type:

None

class noether.core.callbacks.early_stoppers.FixedEarlyStopper(callback_config, **kwargs)¶

Bases: noether.core.callbacks.early_stoppers.base.EarlyStopperBase

Early stopper (training) based on a fixed number of epochs, updates, or samples.

Example config:

- kind: noether.core.callbacks.FixedEarlyStopper
  stop_at_epoch: 10
  name: FixedEarlyStopper

Parameters:

callback_config (FixedEarlyStopperConfig) – The configuration for the callback. See FixedEarlyStopperConfig for available options.
**kwargs – Additional arguments to pass to the parent class.

stop_at_sample¶

stop_at_update¶

stop_at_epoch¶

class noether.core.callbacks.early_stoppers.FixedEarlyStopperConfig(/, **data)¶

Bases: pydantic.BaseModel

Parameters:: data (Any)

kind: str | None = None¶

name: Literal['FixedEarlyStopper'] = None¶

stop_at_sample: int | None = None¶

stop_at_update: int | None = None¶

stop_at_epoch: int | None = None¶

validate_callback_frequency()¶

Ensures that exactly one stop (‘stop_at_*’) is specified

Return type:: FixedEarlyStopperConfig

class noether.core.callbacks.early_stoppers.MetricEarlyStopper(callback_config, **kwargs)¶

Bases: noether.core.callbacks.early_stoppers.base.EarlyStopperBase

Early stopper (training) based on a metric value to be monitored.

Example config:

- kind: noether.core.callbacks.MetricEarlyStopper
  every_n_epochs: 1
  metric_key: loss/val/total
  tolerance: 0.10
  name: MetricEarlyStopper

Parameters:

callback_config (MetricEarlyStopperConfig) – Configuration for the callback. See MetricEarlyStopperConfig for available options including metric key and tolerance.
**kwargs – Additional arguments to pass to the parent class.

metric_key¶

higher_is_better¶

tolerance¶

tolerance_counter = 0¶

best_metric¶

class noether.core.callbacks.early_stoppers.MetricEarlyStopperConfig(/, **data)¶

Bases: noether.core.callbacks.base.CallBackBaseConfig

Internal base class for all registry-based configs.

Provides auto-registration via __init_subclass__. Not meant to be used directly - use specific config base classes instead.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

Parameters:: data (Any)

name: Literal['MetricEarlyStopper'] = None¶

metric_key: str¶: The key of the metric to monitor

tolerance: int¶: The number of times the metric can stagnate before stopping training

classmethod check_tolerance_positive(v)¶

Ensures that tolerance is at least 1.

Parameters:: v (int)
Return type:: int