niftynet.engine.handler_model module

This module implements a model checkpoint loader and writer.

make_model_name(model_dir)

Make the model checkpoint folder. the checkpoint file will be located at model_dir/models/ folder, the filename will start with FILE_PREFIX.

Parameters:model_dir – niftynet model folder Returns:a partial name of a checkpoint file model_dir/model/FILE_PREFIX

class ModelRestorer(model_dir, initial_iter=0, is_training_action=True, vars_to_restore=None, **_unused)

Bases: object

This class handles restoring the model at the beginning of a session.

rand_init_model(_sender, **_unused)

Randomly initialising all trainable variables defined in the default session.

Parameters:

• _sender –
• _unused –

Returns:

restore_model(_sender, **_unused)

Loading checkpoint files as variable initialisations.

Parameters:

• _sender –
• _unused –

Returns:

class ModelSaver(model_dir, save_every_n=0, max_checkpoints=1, is_training_action=True, **_unused)

Bases: object

This class handles iteration events to save the model as checkpoint files.

init_saver(_sender, **_unused)

Initialise a model saver.

Parameters:

• _sender –
• _unused –

Returns:

save_model(_sender, **msg)

Saving the model at the current iteration.

Parameters:

• _sender –
• msg – an iteration message instance

Returns:

save_model_interval(_sender, **msg)

Saving the model according to the frequency of save_every_n.

Parameters:

• _sender –
• msg – an iteration message instance

Returns: