pytorch_lightning.utilities.cloud_io module

pytorch_lightning.utilities.cloud_io.atomic_save(checkpoint, filepath)

Saves a checkpoint atomically, avoiding the creation of incomplete checkpoints.

Parameters

• checkpoint – The object to save. Built to be used with the dump_checkpoint method, but can deal with anything which torch.save accepts.

• filepath (str) – The path to which the checkpoint will be saved. This points to the file that the checkpoint will be stored in.

pytorch_lightning.utilities.cloud_io.cloud_open(path, mode, newline=None)

pytorch_lightning.utilities.cloud_io.is_remote_path(path)

Determine if a path is a local path or a remote path like s3://bucket/path

This should catch paths like s3:// hdfs:// and gcs://

pytorch_lightning.utilities.cloud_io.load(path_or_url, map_location=None)

pytorch_lightning.utilities.cloud_io.makedirs(path)

pytorch_lightning.utilities.cloud_io.modern_gfile()

Check the version number of tensorboard.

Cheking to see if it has the gfile compatibility layers needed for remote file operations