GradientCallback

class radionets.core.callbacks.GradientCallback(num_epochs, validation_data, arch_name, amp_phase)

Bases: Callback

Callback for gradient and prediction tracking.

Parameters:

num_epochsint

Number of training epochs.

validation_datastr or Path

Path to the validation dataset.

arch_namestr

Name of the architecture used for the model.

amp_phasebool

Whether to use amplitude-phase representation.

Attributes Summary

learn
name	Name of the Callback, camel-cased and with 'Callback' removed
order
run
run_train
run_valid

Methods Summary

__call__(event_name)	Call self.{event_name} if it's defined
after_cancel_backward()
after_epoch()
before_backward()

Attributes Documentation

learn = None

name

Name of the Callback, camel-cased and with ‘Callback’ removed

order = 0

run = True

run_train = True

run_valid = True

Methods Documentation

__call__(event_name)

Call self.{event_name} if it’s defined

after_cancel_backward()

after_epoch()

before_backward()