CheckpointSaver¶

class tfsnippet.CheckpointSaver(variables, save_dir, objects=None, filename='checkpoint.dat', max_to_keep=None, save_meta=True, name=None, scope=None)¶

Bases: tfsnippet.utils.reuse.VarScopeObject

Save and restore tf.Variable, ScheduledVariable and CheckpointSavableObject with tf.train.Saver.

Attributes Summary

`filename`	Get the filename of checkpoint files.
`name`	Get the name of this object.
`save_dir`	Get the checkpoint directory.
`save_meta`	Whether or not to save graph meta?
`saver`	Get the TensorFlow saver object.
`variable_scope`	Get the variable scope of this object.

Methods Summary

`latest_checkpoint`()	Get the path of the latest checkpoint file.
`recover_internal_states`()	Restore the internal states of this saver.
`restore`(save_path[, session])	Restore from a checkpoint file.
`restore_latest`([ignore_non_exist, session])	Restore the latest checkpoint file.
`save`([global_step, session])	Save the session to a checkpoint file.

Attributes Documentation

filename¶: Get the filename of checkpoint files.

name¶: Get the name of this object.

save_dir¶: Get the checkpoint directory.

save_meta¶: Whether or not to save graph meta?

saver¶

Get the TensorFlow saver object.

Returns:	The TensorFlow saver object.
Return type:	tf.train.Saver

variable_scope¶: Get the variable scope of this object.

Methods Documentation

latest_checkpoint()¶

Get the path of the latest checkpoint file.

Returns:	The path of the latest checkpoint file, or `None` if no checkpoint file is found.
Return type:	str or None

recover_internal_states()¶: Restore the internal states of this saver.

restore(save_path, session=None)¶

Restore from a checkpoint file.

Parameters:	save_path (str) – Restore from this checkpoint file. session (tf.Session) – Restore the variables into this session. If not specified, restore into the default session.

restore_latest(ignore_non_exist=False, session=None)¶

Restore the latest checkpoint file. :param ignore_non_exist: Whether or not to ignore error if the

latest checkpoint file does not exist?

Parameters:	session (tf.Session) – Restore the variables into this session. If not specified, restore into the default session.
Raises:	`IOError` – If no checkpoint file is found.

save(global_step=None, session=None)¶

Save the session to a checkpoint file.

Parameters:	global_step (int or tf.Tensor) – The global step counter. session (tf.Session) – The session to save. If not specified, select the default session.
Returns:	The path of the saved checkpoint file.
Return type:	str