New Features in April: Run Spell in your AWS Account, PyTorch Model Serving, and Bayesian Hyperparameter Search

It’s April and spring has come to New York City and Spell. We have a bunch of big new features we’re launching this month. Head to spell.run to test them out and tell us what you think!

Spell Cluster Management

Adding a new machine to a Spell-managed cluster

Since we announced Dedicated Clusters in January we’ve been onboarding customers and starting them with our cluster management tool. Based on feedback, we’re rolling out new features and updates to the UI that makes it easier to manage your own dedicated AWS cluster using Spell.

With this month’s release, we now support:

• Model serving. Serve both Tensorflow and PyTorch models from your own machines.
• Shared Jupyter workspaces. Only a few months old and already one of our most popular features, you can launch shared Jupyter workspaces on your own cluster.
• SpellFS. ls, cp, mount, and link work in your own cluster (and are easier to navigate than AWS’s default file system).

PyTorch Model Serving

In addition to our Tensorflow model serving, we now support PyTorch model serving for users on our Teams plan.

Interested in model serving with Spell? Email us at support@spell.run or request a demo through this form.

Bayesian Hyperparameter Search

You may have seen our post last week about our new Bayesian Hyperparameter search tool. Bayesian optimization gives a more informed search than the grid or random search, which are two popular hyperparameter search methods we also support. An informed search means that you’ll get to a result faster, and that the result will be better, too.

Watching the results of a Bayesian hyperparameter search in the web console

Compress and Upload

Large amounts of data are the foundation of any machine learning experiment. And waiting for large datasets to upload is a waste of time. Now you can compress and upload large files with the -c parameter:

$ spell upload -c my_really_big_files

💡 Tip of the month

Ever find yourself running the same command more than once? With the --idempotent parameter, you can now save time when re-running runs.

Adding--idempotent to a run will search for a previous run that is identical to what you requested (i.e., identical pip, apt, command, commit hash, environment variables, etc.). And if it finds a match it returns the previous run in lieu of re-running.

Stay tuned

We’ve got a lot more exciting features in the works that we’ll be announcing soon. Be the first to know by following us on Twitter or here on Medium!