ray.train.lightning.LightningCheckpoint

class ray.train.lightning.LightningCheckpoint(*args, **kwargs)

Bases: ray.train.torch.torch_checkpoint.TorchCheckpoint

A Checkpoint with Lightning-specific functionality.

LightningCheckpoint only support file based checkpoint loading. Create this by calling LightningCheckpoint.from_directory(ckpt_dir), LightningCheckpoint.from_uri(uri) or LightningCheckpoint.from_path(path)

LightningCheckpoint loads file named model under the specified directory.

Examples:

from ray.train.lightning import LightningCheckpoint

# Suppose we saved a checkpoint in "./checkpoint_000000/model":
# Option 1 (Preferred): Load from the checkpoint file
checkpoint = LightningCheckpoint.from_path(
    path="./checkpoint_00000/model"
)

# Option 2: Load from a directory
checkpoint = LightningCheckpoint.from_directory(
    path="./checkpoint_00000/"
)

# Suppose we saved a checkpoint in an S3 bucket:
# Option 3: Load from URI
checkpoint = LightningCheckpoint.from_uri(
    path="s3://path/to/checkpoint/directory/"
)

PublicAPI (alpha): This API is in alpha and may change before becoming stable.

Methods

as_directory()	Return checkpoint directory path in a context.
from_bytes(data)	Create a checkpoint from the given byte string.
from_checkpoint(other)	Create a checkpoint from a generic ray.air.checkpoint.Checkpoint.
from_dict(data)	Create checkpoint object from dictionary.
from_directory(path)	Create checkpoint object from directory.
from_model(model, *[, preprocessor])	Create a Checkpoint that stores a Torch model.
from_path(path, *[, preprocessor])	Create a ray.air.lightning.LightningCheckpoint from a checkpoint file.
from_state_dict(state_dict, *[, preprocessor])	Create a Checkpoint that stores a model state dictionary.
from_uri(uri)	Create checkpoint object from location URI (e.g.
get_internal_representation()	Return tuple of (type, data) for the internal representation.
get_model(model_class, ...)	Retrieve the model stored in this checkpoint.
get_preprocessor()	Return the saved preprocessor, if one exists.
set_preprocessor(preprocessor)	Saves the provided preprocessor to this Checkpoint.
to_bytes()	Return Checkpoint serialized as bytes object.
to_dict()	Return checkpoint data as dictionary.
to_directory([path])	Write checkpoint data to directory.
to_uri(uri)	Write checkpoint data to location URI (e.g.

Attributes

path	Return path to checkpoint, if available.
uri	Return checkpoint URI, if available.