Checkpointing

Lightning provides functions to save and load checkpoints.

Checkpointing your training allows you to resume a training process in case it was interrupted, fine-tune a model or use a pre-trained model for inference without having to retrain the model.

Checkpoint Contents

A Lightning checkpoint has everything needed to restore a training session including:

• 16-bit scaling factor (if using 16-bit precision training)

• Current epoch

• Global step

• LightningModule’s state_dict

• State of all optimizers

• State of all learning rate schedulers

• State of all callbacks (for stateful callbacks)

• State of datamodule (for stateful datamodules)

• The hyperparameters used for that model if passed in as hparams (Argparse.Namespace)

• State of Loops (if using Fault-Tolerant training)

Individual Component States

Each component can save and load its state by implementing the PyTorch state_dict, load_state_dict stateful protocol. For details on implementing your own stateful callbacks and datamodules, refer to the individual docs pages at callbacks and datamodules.

Operating on Global Checkpoint Component States

If you need to operate on the global component state (i.e. the entire checkpoint dictionary), you can read/add/delete/modify custom states in your checkpoints before they are being saved or loaded. For this you can override on_save_checkpoint() and on_load_checkpoint() in your LightningModule or on_save_checkpoint() and on_load_checkpoint() methods in your Callback.

Checkpoint Saving

Automatic Saving

Lightning automatically saves a checkpoint for you in your current working directory, with the state of your last training epoch. This makes sure you can resume training in case it was interrupted.

To change the checkpoint path pass in:

# saves checkpoints to '/your/path/to/save/checkpoints' at every epoch end
trainer = Trainer(default_root_dir="/your/path/to/save/checkpoints")

You can retrieve the checkpoint after training by calling:

checkpoint_callback = ModelCheckpoint(dirpath="my/path/", save_top_k=2, monitor="val_loss")
trainer = Trainer(callbacks=[checkpoint_callback])
trainer.fit(model)
checkpoint_callback.best_model_path

Disabling Checkpoints

You can disable checkpointing by passing:

trainer = Trainer(enable_checkpointing=False)

Manual Saving

You can manually save checkpoints and restore your model from the checkpointed state using save_checkpoint() and load_from_checkpoint().

model = MyLightningModule(hparams)
trainer.fit(model)
trainer.save_checkpoint("example.ckpt")
new_model = MyLightningModule.load_from_checkpoint(checkpoint_path="example.ckpt")

Manual Saving with Distributed Training Strategies

Lightning also handles strategies where multiple processes are running, such as DDP. For example, when using the DDP strategy our training script is running across multiple devices at the same time. Lightning automatically ensures that the model is saved only on the main process, whilst other processes do not interfere with saving checkpoints. This requires no code changes as seen below:

trainer = Trainer(strategy="ddp")
model = MyLightningModule(hparams)
trainer.fit(model)
# Saves only on the main process
trainer.save_checkpoint("example.ckpt")

Not using save_checkpoint() can lead to unexpected behavior and potential deadlock. Using other saving functions will result in all devices attempting to save the checkpoint. As a result, we highly recommend using the Trainer’s save functionality. If using custom saving functions cannot be avoided, we recommend using the rank_zero_only() decorator to ensure saving occurs only on the main process. Note that this will only work if all ranks hold the exact same state and won’t work when using model parallel distributed strategies such as deepspeed or sharded training.

Checkpointing Hyperparameters

The Lightning checkpoint also saves the arguments passed into the LightningModule init under the "hyper_parameters" key in the checkpoint.

class MyLightningModule(LightningModule):
    def __init__(self, learning_rate, *args, **kwargs):
        super().__init__()
        self.save_hyperparameters()


# all init args were saved to the checkpoint
checkpoint = torch.load(CKPT_PATH)
print(checkpoint["hyper_parameters"])
# {"learning_rate": the_value}

---

Checkpoint Loading

To load a model along with its weights and hyperparameters use the following method:

model = MyLightningModule.load_from_checkpoint(PATH)

print(model.learning_rate)
# prints the learning_rate you used in this checkpoint

model.eval()
y_hat = model(x)

But if you don’t want to use the hyperparameters saved in the checkpoint, pass in your own here:

class LitModel(LightningModule):
    def __init__(self, in_dim, out_dim):
        super().__init__()
        self.save_hyperparameters()
        self.l1 = nn.Linear(self.hparams.in_dim, self.hparams.out_dim)

you can restore the model like this

# if you train and save the model like this it will use these values when loading
# the weights. But you can overwrite this
LitModel(in_dim=32, out_dim=10)

# uses in_dim=32, out_dim=10
model = LitModel.load_from_checkpoint(PATH)

# uses in_dim=128, out_dim=10
model = LitModel.load_from_checkpoint(PATH, in_dim=128, out_dim=10)

Restoring Training State

If you don’t just want to load weights, but instead restore the full training, do the following:

model = LitModel()
trainer = Trainer()

# automatically restores model, epoch, step, LR schedulers, apex, etc...
trainer.fit(model, ckpt_path="some/path/to/my_checkpoint.ckpt")

---

Conditional Checkpointing (ModelCheckpoint)

The ModelCheckpoint callback allows you to configure when/which/what/where checkpointing should happen. It follows the normal Callback hook structure so you can hack it around/override its methods for your use-cases as well. Following are some of the common use-cases along with the arguments you need to specify to configure it:

How does it work?

ModelCheckpoint helps cover the following cases from WH-Family:

When

• When using iterative training which doesn’t have an epoch, you can checkpoint at every N training steps by specifying every_n_training_steps=N.

• You can also control the interval of epochs between checkpoints using every_n_epochs between checkpoints, to avoid slowdowns.

• You can checkpoint at a regular time interval using train_time_interval argument independent of the steps or epochs.

• In case you are monitoring a training metrics, we’d suggest using save_on_train_epoch_end=True to ensure the required metric is being accumulated correctly for creating a checkpoint.

Which

• You can save the last checkpoint when training ends using save_last argument.

• You can save top-K and last-K checkpoints by configuring the monitor and save_top_k argument.

from pytorch_lightning.callbacks import ModelCheckpoint


# saves top-K checkpoints based on "val_loss" metric
checkpoint_callback = ModelCheckpoint(
    save_top_k=10,
    monitor="val_loss",
    mode="min",
    dirpath="my/path/",
    filename="sample-mnist-{epoch:02d}-{val_loss:.2f}",
)

# saves last-K checkpoints based on "global_step" metric
# make sure you log it inside your LightningModule
checkpoint_callback = ModelCheckpoint(
    save_top_k=10,
    monitor="global_step",
    mode="max",
    dirpath="my/path/",
    filename="sample-mnist-{epoch:02d}-{global_step}",
)

• You can customize the checkpointing behavior to monitor any quantity of your training or validation steps. For example, if you want to update your checkpoints based on your validation loss:

from pytorch_lightning.callbacks import ModelCheckpoint


class LitAutoEncoder(LightningModule):
    def validation_step(self, batch, batch_idx):
        x, y = batch
        y_hat = self.backbone(x)

        # 1. calculate loss
        loss = F.cross_entropy(y_hat, y)

        # 2. log val_loss
        self.log("val_loss", loss)


# 3. Init ModelCheckpoint callback, monitoring "val_loss"
checkpoint_callback = ModelCheckpoint(monitor="val_loss")

# 4. Add your callback to the callbacks list
trainer = Trainer(callbacks=[checkpoint_callback])

What

• By default, the ModelCheckpoint callback saves model weights, optimizer states, etc., but in case you have limited disk space or just need the model weights to be saved you can specify save_weights_only=True.

Where

• It gives you the ability to specify the dirpath and filename for your checkpoints. Filename can also be dynamic so you can inject the metrics that are being logged using log().

from pytorch_lightning.callbacks import ModelCheckpoint


# saves a file like: my/path/sample-mnist-epoch=02-val_loss=0.32.ckpt
checkpoint_callback = ModelCheckpoint(
    dirpath="my/path/",
    filename="sample-mnist-{epoch:02d}-{val_loss:.2f}",
)

The ModelCheckpoint callback is very robust and should cover 99% of the use-cases. If you find a use-case that is not configured yet, feel free to open an issue with a feature request on GitHub and the Lightning Team will be happy to integrate/help integrate it.

---

Customize Checkpointing

Warning

The Checkpoint IO API is experimental and subject to change.

Lightning supports modifying the checkpointing save/load functionality through the CheckpointIO. This encapsulates the save/load logic that is managed by the Strategy. CheckpointIO is different from on_save_checkpoint() and on_load_checkpoint() methods as it determines how the checkpoint is saved/loaded to storage rather than what’s saved in the checkpoint.

Built-in Checkpoint IO Plugins

Built-in Checkpoint IO Plugins

Plugin	Description
TorchCheckpointIO	CheckpointIO that utilizes torch.save() and torch.load() to save and load checkpoints respectively, common for most use cases.
XLACheckpointIO	CheckpointIO that utilizes xm.save() to save checkpoints for TPU training strategies.

Custom Checkpoint IO Plugin

CheckpointIO can be extended to include your custom save/load functionality to and from a path. The CheckpointIO object can be passed to either a Trainer directly or a Strategy as shown below:

from pytorch_lightning import Trainer
from pytorch_lightning.callbacks import ModelCheckpoint
from pytorch_lightning.plugins import CheckpointIO
from pytorch_lightning.strategies import SingleDeviceStrategy


class CustomCheckpointIO(CheckpointIO):
    def save_checkpoint(self, checkpoint, path, storage_options=None):
        ...

    def load_checkpoint(self, path, storage_options=None):
        ...

    def remove_checkpoint(self, path):
        ...


custom_checkpoint_io = CustomCheckpointIO()

# Either pass into the Trainer object
model = MyModel()
trainer = Trainer(
    plugins=[custom_checkpoint_io],
    callbacks=ModelCheckpoint(save_last=True),
)
trainer.fit(model)

# or pass into Strategy
model = MyModel()
device = torch.device("cpu")
trainer = Trainer(
    strategy=SingleDeviceStrategy(device, checkpoint_io=custom_checkpoint_io),
    callbacks=ModelCheckpoint(save_last=True),
)
trainer.fit(model)

Note

Some strategies like DeepSpeedStrategy do not support custom CheckpointIO as checkpointing logic is not modifiable.

---

Managing Remote Filesystems

Lightning supports saving and loading checkpoints from a variety of filesystems, including local filesystems and several cloud storage providers.

Check out Remote Filesystems document for more info.