ClearML tracks models (input models, output models and model snapshots), and other objects in several formats as experiment artifacts. Once ClearML stores a model in ClearML Server, its configuration and label enumeration are editable. You and your teammates can run experiments with different, stored initial weights models, or work with snapshots.

Artifacts are stored in the ClearML Hosted Service (or self-hosted ClearML Server). Other artifact objects can be uploaded and dynamically tracked, or uploaded without tracking. You can configure ClearML for the storage you use, see configuring ClearML for artifact storage.

See the artifacts and model upload examples for Keras, PyTorch, and TensorFlow.

Next Steps