qkeras.callbacks

Functions

create_file_writer(log_dir)	Create a writer object with .as_default() context manager.
scalar(name, data, step)	Write a scalar under the active default writer.

Classes

QNoiseScheduler(start, finish[, freq_type, ...])

Schedules the gradual quantization noise training for each step (or epoch).

class qkeras.callbacks.QNoiseScheduler(start, finish, freq_type='epoch', update_freq=1, initial_step_or_epoch=0, exponent=3.0, use_ste=True, log_dir=None)

Bases: Callback

Schedules the gradual quantization noise training for each step (or epoch).

It updates the qnoise_factor in the quantizers to gradually introduce the quantization noise during training.

The idea was adopted from “https://arxiv.org/pdf/1903.01061.pdf”

calculate_qnoise_factor(freq)

Returns calculated qnoise_factor based on the current step (epoch) and the schedule parameters.

Parameters:

freq – The current step (or epoch) to calculate the qnoise_factor.

Returns:

calculated qnoise_factor.

Return type:

qnoise_factor

get_quantizers(model)

Returns a list of quantizers with qnoise_factor in the model.

Parameters:

model – model to get a list of quantizers with qnoise_factor.

Returns:

A list of quantizers with the qnoise_factor variable.

on_epoch_begin(epoch, logs=None)

Called at the start of an epoch.

Subclasses should override for any actions to run. This function should only be called during TRAIN mode.

Parameters:

• epoch – Integer, index of epoch.

• logs – Dict. Currently no data is passed to this argument for this method but that may change in the future.

on_epoch_end(epoch, logs=None)

Called at the end of an epoch.

Subclasses should override for any actions to run. This function should only be called during TRAIN mode.

Parameters:

• epoch – Integer, index of epoch.

• logs – Dict, metric results for this training epoch, and for the validation epoch if validation is performed. Validation result keys are prefixed with val_. For training epoch, the values of the Model’s metrics are returned. Example: {‘loss’: 0.2, ‘accuracy’: 0.7}.

on_train_batch_begin(batch, logs=None)

Called at the beginning of a training batch in fit methods.

Subclasses should override for any actions to run.

Note that if the steps_per_execution argument to compile in Model is set to N, this method will only be called every N batches.

Parameters:

• batch – Integer, index of batch within the current epoch.

• logs – Dict. Currently no data is passed to this argument for this method but that may change in the future.

on_train_begin(logs=None)

Called at the beginning of training.

Subclasses should override for any actions to run.

Parameters:

logs – Dict. Currently no data is passed to this argument for this method but that may change in the future.

set_qnoise_factor(quantizer, qnoise_factor)

Set self.qnoise_factor and update the qnoise_factor of the quantizer.

set_quantizers()

Set quantizers to update the qnoise_factor.

This must be called before building the quantizers.

update_qnoise_factor(freq)

Update the qnoise_factor of the model.

Parameters:

freq – The current step (epoch) to calculate the qnoise_factor.

qkeras.callbacks.create_file_writer(log_dir)

Create a writer object with .as_default() context manager.

Parameters:

log_dir (str)

qkeras.callbacks.scalar(name, data, step)

Write a scalar under the active default writer.

Parameters:

• name (str)

• step (int)