CABS

class cockpit.quantities.CABS(track_schedule, verbose=False)[source]

CABS Quantity class for the suggested batch size using the CABS criterion.

CABS uses the current learning rate and variance of the stochastic gradients to suggest an optimal batch size.

Only applies to SGD without momentum.

Note: Proposed in

  • Balles, L., Romero, J., & Hennig, P., Coupling adaptive batch sizes with learning rates (2017).

Initialization sets the tracking schedule & creates the output dict.

Parameters

  • track_schedule (callable) – Function that maps the global_step to a boolean, which determines if the quantity should be computed.

  • verbose (bool, optional) – Turns on verbose mode. Defaults to False.