Catalyst

DVCLive allows you to easily add experiment tracking capabilities to your Catalyst projects.

About Catalyst

Catalyst is a PyTorch framework for Deep Learning Research and Development. It focuses on reproducibility, rapid experimentation, and codebase reuse.

Usage

To start using DVCLive you just need to add a few lines to your training code in any Catalyst project.

You just need to add the DvcLiveCallback to the callbacks list passed to your Runner:

+from dvclive.catalyst import DvcLiveCallback

. . .

runner.train(
    model=model,
    criterion=criterion,
    optimizer=optimizer,
    loaders=loaders,
-    num_epochs=2)
+    num_epochs=2,
+    callbacks=[DvcLiveCallback()])

This will generate the metrics logs and summaries as described in the Get Started.

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

Parameters

• model_file - (None by default) - The name of the file where the model will be saved at the end of each step.

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

Examples

• Using model_file.

from dvclive.catalyst import DvcLiveCallback

runner.train(
    model=model,
    criterion=criterion,
    optimizer=optimizer,
    loaders=loaders,
    num_epochs=2,
    callbacks=[DvcLiveCallback("model.pth")])

• Using **kwargs to customize Live.

from dvclive.catalyst import DvcLiveCallback

runner.train(
    model=model,
    criterion=criterion,
    optimizer=optimizer,
    loaders=loaders,
    num_epochs=2,
    callbacks=[
      DvcLiveCallback(path="custom_path", summary=False)])