<function save_npz>, filename='snapshot_iter_{.updater.iteration}')

Returns a trainer extension to take snapshots of the trainer.

This extension serializes the trainer object and saves it to the output directory. It is used to support resuming the training loop from the saved state.

This extension is called once per epoch by default. To take a snapshot at a different interval, a trigger object specifying the required interval can be passed along with this extension to the extend() method of the trainer.

The default priority is -100, which is lower than that of most built-in extensions.


This extension first writes the serialized object to a temporary file and then rename it to the target file name. Thus, if the program stops right before the renaming, the temporary file might be left in the output directory.

  • savefun – Function to save the trainer. It takes two arguments: the output file path and the trainer object.
  • filename (str) – Name of the file into which the trainer is serialized. It can be a format string, where the trainer object is passed to the str.format() method.