Release Notes 3.0 (October 2024)
MLflow Integration - Phase 1
We're excited to introduce MLflow integration to enhance your machine learning workflows in Code Ocean. This includes an integrated MLflow tracking server that benefits from all the inherent security and access permissions of the Code Ocean platform.
With this, users can:
Create MLflow experiments that will track runs from Capsules and Pipelines
Monitor and manage experiments within the integrated MLflow tracking server
Register models for lifecycle management, version control, and downstream workflows
Models
Just like Data Assets, Models have their own dashboard and come with full lineage and provenance. They also benefit from the same core principles for reproducibility that are built into the platform.
Dark Mode
Users can now choose an alternative visual experience designed to reduce eye strain and enhance usability. From the navigation sidebar, choose between light or dark mode, or follow your computer's settings.
Improvements to Search
Search capabilities across Code Ocean have been enhanced to make finding resources faster and improve accuracy in search results. These updates ensure that you can quickly locate the assets and information you need across your workflows.
To further enhance search in the UI and via the Public API, Release and non-Release Capsules/Pipelines have been separated into distinct line items.
New Dashboard Filters
New dashboard filters are designed to improve your user experience and help you quickly find the assets you need.
Easily filter Capsules, Pipelines, Data, and Models based on criteria specific to each dashboard, such as Source, Type, Release Status, and more. Filter sections use AND logic, making it easier for you to find exactly what you're looking for.
Search Timeline by & Copy Computation ID
Two enhancements have been added to make it easier to manage and track your computational workflows:
Search by Computation ID: You can now search the Capsule and Pipeline Timeline by computation ID, helping you to quickly locate specific runs in the Timeline.
Copy Computation ID: A new option has been added to the Timeline computation run menu, allowing you to quickly copy computation IDs with 1-click. This streamlines sharing and referencing runs across your projects.
Dedicated Resources - Additional EC2 Instance Types
We have expanded support for additional EC2 instances, now to include all compatible x86 instance types available in your AWS region. This enhancement provides users with more options to meet diverse workload requirements, ensuring optimal performance and flexibility in their cloud environment.
Backup Stack - v1.2.1 Available
Version 1.2.1 of the backup stack is now available. This new version includes backup S3 buckets for MLflow and Packages.
If you have a backup stack configured, you must upgrade the backup stack to v1.2.1 prior to upgrading the linked primary deployment to v3.0. You also must add the ARNs of the new Mlflow and Packages backup S3 buckets to their respective parameters in the CloudFormation template when upgrading the primary deployment. More information and the latest backup stack CloudFormation template can be found in the Admin Guide here.
Public API - Extended
Expose exit_code in Get Computation Endpoint Response
The Computation exit_code is now part of the Get Computation response so users can have a reliable indicator of the final status of the computation. The exit_code value will be 0 when the code runs successfully and non zero when it doesn’t. Any error in a Pipeline will have an exit code of 1.
Rename Computation
A new endpoint has been introduced to allow users to rename a run via the public API.
Admin and Owner Ability to Create Data Asset from another User's Run
Admins and Owners of Capsules now have the ability to create a Data Asset from another user's run via the API. This matches the functionality currently available in the Code Ocean UI.
Last updated