Method save_checkpoints

caffe2/python/checkpoint.py:789–816

Triggers operation to save checkpoints This method will trigger the Save ops to serialize and persist the blobs present in the global workspaace. Args: epoch: An integer. The checkpoint epoch-id that we are saving. session: A Session object to execut

(self, epoch, session)

        return result

    def save_checkpoints(self, epoch, session):
        """Triggers operation to save checkpoints

        This method will trigger the Save ops to serialize and persist the
        blobs present in the global workspaace.

        Args:
            epoch: An integer. The checkpoint epoch-id that we are saving.
            session: A Session object to execute the save ops.

        Raises:
            ValueError: When the checkpoint manager is invalid.
        """
        if not self.checkpoint_manager:
            raise ValueError('Checkpoint manager is None')
        try:
            is_accessible = self.checkpoint_manager.cp_accessible(epoch=None)
            if is_accessible:
                logger.info('Saving checkpoints for epoch {}'.format(epoch))
                session.run(self.checkpoint_manager.save(epoch))
                self.checkpoint_manager.write_checkpoint_metadata(epoch)
                logger.info('Checkpoints saved')
                self.checkpoint_manager.report_checkpoint_stats('checkpoint_save')
            else:
                logger.warning("Checkpoint files cannot be accessed!")
        except Exception as ex:
            logger.warning("Unable to write checkpoint for epoch {}. Error={}".
                            format(epoch, ex))


def epoch_limiter(job, num_epochs):