PureML provides the ability to log a range of metadata related to models and datasets. The specific types of metadata that can be logged via the PureML package and how they are represented in the PureML app will depend on the data type and format being used.
For model metadata, you can log key-value pairs of the model-building parameters and key-value pairs related to the performance of a model version.
- Parameters: log key-value pairs of the model-building parameters for a version
- Metrics : log key-value pairs related to the performance of a model version
You can register artifacts, such as files, through the artifacts API to track them across different model versions. Once you’ve registered an artifact, all the corresponding metadata will be uploaded to the PureML model.
You can register a variety of data types and file formats to a PureML model, including tabular data, arrays, visualizations, images, audio, and video. Registered files can be viewed directly in the PureML web interface. Note that certain data types and formats may not yet be fully supported by the platform, but we are constantly working on expanding our capabilities and supporting additional formats in the future.
- Video - Coming Soon!!
- Audio - Coming Soon!!
- Images - Coming Soon!!
- Visualizations - Coming Soon!!
- Arrays - Coming Soon!!
- Tabular data - Coming Soon!!