NPUCheckpointSaverHook Constructor

Applicability

Product	Supported
Atlas A3 training products/Atlas A3 inference products	√
Atlas A2 training products/Atlas A2 inference products	√
Atlas 200I/500 A2 inference products	☓
Atlas inference products	☓
Atlas training products	√

Description

Constructs an object of class NPUCheckpointSaverHook, which is used to save the checkpoint file. The NPUCheckpointSaverHook class inherits the CheckpointSaverHook class and can call the native APIs of the base class to record model information during training.

Prototype

class NPUCheckpointSaverHook(basic_session_run_hooks.CheckpointSaverHook):
    def __init__(self,
                 checkpoint_dir,
                 save_secs=None,
                 save_steps=None,
                 saver=None,
                 checkpoint_basename="model.ckpt",
                 scaffold=None,
                 listeners=None)

Parameters

Parameter	Input/Output	Description
checkpoint_dir	Input	Checkpoint file directory
save_secs	Input	Interval (in seconds) for saving the checkpoint file
save_steps	Input	Interval (in steps) for saving the checkpoint file
saver	Input	Saver object
checkpoint_basename	Input	Basename of the checkpoint file
scaffold	Input	Scaffold of the saver object
listeners	Input	Example of the CheckpointSaverListener subclass, for saving the checkpoint file

Returns

An object of the NPUCheckpointSaverHook class

Restrictions

When NPUEstimator is used and iteration_per_loop is set to a value greater than 1, the hook may not take effect.

Example

from npu_bridge.npu_init import *
checkpoint_hook = NPUCheckpointSaverHook(checkpoint_dir='./ckpt', save_steps=2000)
...
mnist_classifier.train(   
   input_fn=train_input_fn,
    steps=2000,
    hooks=[checkpoint_hook])

Parent topic: npu_bridge.estimator.npu.npu_hook