Core Concepts
This page is a high-level overview of Spell's most important features. For more detailed discussions of specific features, refer to the corresponding sections of the User Guide.
Runs
A run is a single instance of a computational job executed on Spell.
Runs take your code, data, and training environment as input, execute the training job on the cloud, and save the outputs to object storage. We do a lot of work behind the scenes to make runs easy to use and ergonomic.
To learn more about runs refer to the Run Overview.
Workspaces
A workspace is a JupyterLab instance running on the cloud.
Workspaces provide a flexible (CPU or GPU) cloud-native development environment. We manage the data storage and compute environment for you so that you can focus on the code.
To learn more about workspaces refer to the Workspace Overview.
Projects and Experiments
Spell projects allow you to group your runs into meaningful categories. They allow you to create a summary view highlighting key metrics over time.
You can further subdivide your projects into experiments. Experiments allow you to generate reports on specific aspects of your projects.
These features work together to make it easy to share and report project state with your collaborators and project stakeholders.
To learn more, refer to the Project Overview and Experiment Overview pages in the docs.
Hyperparameter Searches
Spell's hyperparameter search features allows you to perform parallelized hyperparameter searches across a pool of cloud machines. We manage the worker machines and handle partitioning your search across a set of runs for you.
To learn more about hyperparameter search see the Hyperparameter Searches guide.
Model Servers
Model servers allow you to serve machine learning models using a managed Kubernetes cluster running on Spell. Model servers allow you to use one platform for both your model training and model serving needs.
To learn more about model servers see the Model Servers guide.
Other Features
That concludes our high-level tour of Spell! This list is not exhaustive, we support many other features like:
- Workflows
- Distributed runs
- TensorBoard support
- Integration with Weights & Biases
- Private machine types
- And more...
To learn more about Spell, check out our blog, CLI quickstart, and/or Python quickstart. For details on specific Spell features, refer to the corresponding section of the docs.