Skip to content
Edit on GitHub

PyTorch Lightning

DVCLive allows you to add experiment tracking capabilities to your PyTorch Lightning projects.

Usage

To start using DVCLive, add a few lines to your training code in any PyTorch Lightning project.

Pass the DvcLiveLogger to your Trainer:

+from dvclive.lightning import DvcLiveLogger

...
 dvclive_logger = DvcLiveLogger()

 trainer = Trainer(
+       logger=dvclive_logger,
    )
 trainer.fit(model)

The history of each {metric} will be stored in:

{Live_dir}/scalars/{split}/{iter_type}/{metric}.tsv

Where:

  • {Live.dir} is the dir attribute of Live.
  • {split} can be either train or eval.
  • {iter_type} can be either epoch or step.
  • {metric} is the name provided by the framework.

Without requiring additional modifications to your training code, you can use DVCLive alongside DVC. See DVCLive with DVC for more info.

Parameters

  • run_name - (None by default) - Name of the run, used in PyTorch Lightning to get version.

  • prefix - (None by default) - string that adds to each metric name.

  • experiment - (None by default) - Live object to be used instead of initializing a new one.

  • **kwargs - Any additional arguments will be passed to Live.

Examples

  • Using **kwargs to customize Live.
from dvclive.lightning import DvcLiveLogger

dvclive_logger = DvcLiveLogger(
    path='my_logs_path'
)
trainer = Trainer(
    logger=dvclive_logger,
)
trainer.fit(model)

By default, PyTorch Lightning creates a directory to store checkpoints using the logger's name (DvcLiveLogger). You can change the checkpoint path or disable checkpointing at all as described in the PyTorch Lightning documentation

Content

🐛 Found an issue? Let us know! Or fix it:

Edit on GitHub

Have a question? Join our chat, we will help you:

Discord Chat