nemos.callbacks.StochasticFitSummary#

class nemos.callbacks.StochasticFitSummary(pass_idx=None, batch_idx=None, n_passes=0, should_stop=False, stop_reason='')[source]#

Bases: object

Post-fit summary of a stochastic training run.

Stored on fitted models as stochastic_fit_summary_ after stochastic_fit completes. This summary is runtime state and is not serialized by save_params().

In machine-learning terminology, a pass over the data is what is usually called an epoch; NeMoS says “pass” to avoid clashing with pynapple’s recording epochs.

Parameters:

  • pass_idx (int | None) – Final pass index reached by the training loop.

  • batch_idx (int | None) – Final batch index reached within the last pass.

  • n_passes (int) – Total number of passes requested for the run.

  • should_stop (bool) – Whether a callback requested early stopping.

  • stop_reason (str) – Human-readable stop reason, if any.

Attributes

batch_idx

n_passes

pass_idx

should_stop

stop_reason

__init__(pass_idx=None, batch_idx=None, n_passes=0, should_stop=False, stop_reason='')#
Parameters:

  • pass_idx (int | None)

  • batch_idx (int | None)

  • n_passes (int)

  • should_stop (bool)

  • stop_reason (str)

Return type:

None

Methods

__init__([pass_idx, batch_idx, n_passes, ...])

batch_idx: int | None = None#
n_passes: int = 0#
pass_idx: int | None = None#
should_stop: bool = False#
stop_reason: str = ''#