edflow.iterators.template_iterator module

Summary

Classes:

TemplateIterator

A specialization of PyHookedModelIterator which adds reasonable default behaviour.

Reference

class edflow.iterators.template_iterator.TemplateIterator(*args, **kwargs)

Bases: edflow.iterators.model_iterator.PyHookedModelIterator

A specialization of PyHookedModelIterator which adds reasonable default behaviour. Subclasses should implement save, restore and step_op.

__init__(*args, **kwargs)

Constructor.

Parameters

• model (object) – Model class.

• num_epochs (int) – Number of times to iterate over the data.

• hooks (list) – List containing Hook instances.

• hook_freq (int) – Frequency at which hooks are evaluated.

• bar_position (int) – Used by tqdm to place bars at the right position when using multiple Iterators in parallel.

initialize(checkpoint_path=None)

step_ops()

Defines ops that are called at each step.

Returns

Return type

The operation run at each step.

save(checkpoint_path)

Save state to checkpoint path.

restore(checkpoint_path)

Restore state from checkpoint path.

step_op(model, **kwargs)

Actual step logic. By default, a dictionary with keys ‘train_op’, ‘log_op’, ‘eval_op’ and callable values is expected. ‘train_op’ should update the model’s state as a side-effect, ‘log_op’ will be logged to the project’s train folder. It should be a dictionary with keys ‘images’ and ‘scalars’. Images are written as png’s, scalars are written to the log file and stdout. Outputs of ‘eval_op’ are written into the project’s eval folder to be evaluated with edeval.