- Release Notes
- Before you begin
- Getting started
- Installing AI Center
- Migration and upgrade
- Projects
- Datasets
- Data Labeling
- ML packages
- Out of the box packages
- Pipelines
- ML Skills
- ML Logs
- Document UnderstandingTM in AI Center
- AI Center API
- How to
- Licensing
- Basic Troubleshooting Guide
2023.10
Release date: 3 November 2023
The queuing mechanism is now available. It manages GPU usage requests, monitors GPU status, and executes the requests when a GPU becomes available.
PUT: /v2/mlskills/stop/{mlSkillId}
is now part of AIC.MLSkills.Edit
. For more information, see:
You can now opt-in for encrypting datasets at rest. For more information, check the Managing Datasets page.
You can now download a report for ML Skills and pipelines. This report gathers all the necessary information required to debug an issue, including account ID, tenant ID, AI Units, and the respective package, pipeline, or ML Skill information.
We recommend attaching this report when submitting an issue for easier troubleshooting.
This release brings a new update to the Multilabel Text Classification model with the following changes:
- Model is moved to Python 3.9.
- Fixed Python dependency.
The AI Units per tenant feature is now available, and you can allocate and track AI Units consumption at tenant level. To benefit from this change, redeploy any running skills. If there are any running pipelines, they will still consume AI Units at the account level until the changes occur.
By default, each tenant is allocated 0 AI units, and all AI units are consumed from the account pool. To make use of this feature, assign AI units limits to each tenant. The tenant will start consuming AI units from its tenant pool until it reaches 0, and at this point new pipelines and new skill deployments will fail. Existing deployments will continue to run.
The Display Name field is now available, where you can add a display name for your pipelines.
There are two new APIs:
- Upload custom model
- Cancel running pipeline
For the full list of available APIs, see the AI Center Permissions list.
- Field listing in Swagger API documentation is improved.
- The Time based interface from the Create new pipeline run page is improved.
- Project creation is retried in case of failure.
- An error is now displayed when trying to send files with Validation failed status to Action Center.
- Model version list sorting is improved. The models are now sorted based on custom version first. Among the same custom version, they are sorted based on the auto-incremented version.
- The character length for the user display name is now 128 characters.
- The size of the
userName
andemail
fields is increased to 256 characters. - Validation is added for the required fields in the Out of Box labeling template.
- Streaming logs are now disabled for community tenants.
- The Send to labelers button from the Labels tab in the Data Labeling window is now called Go to Action Center. You can click this button to go straight to the Data Labeling search filter in Action Center pending tasks.
- The Export dataset button from the Export tab in the Data Labeling window is now called Export files to dataset. A new Go to dataset button is available on the same page.
- You can now check the failed validation error of a label. To do so, hover over the Validation Failed status of a label from the Labels tab in the Data Labeling window.
- The current configuration is now displayed in the Configure tab, from the Data Labeling window.
- You can now access the original data in the labeling output.
- You can no longer add a value lower than 0.5 for CPU when creating a new ML Skill.
- You can now choose if an ML Skill is private or public when creating the ML Skill.
- We implemented several accessibility improvements.
- Fixed an issue related to the GPU field when viewing ML Skills details.
- Fixed an issue causing the MultiLingual Text Classifier model to fail when trained on a GPU. This issue was fixed with an update to the model (23.9.0).
- Fixed an issue where you could edit unrestricted projects without having Project View permissions.
- The role assignment is no longer displayed in the Role assignment page, when deleting a role.
- Several security and accessibility issues have been fixed.
- When upgrading from 2023.4.3 to 2023.10, you run into issues with provisioning AI Center. The system shows the following exception,
and the tenant creation fails:
"exception":"sun.security.pkcs11.wrapper.PKCS11Exception: CKR_KEY_SIZE_RANGE
- In certain situations, the
Out of the Box Packages
installer can fail. If this happens, some ML Package versions will be missing in Document UnderstandingTM. To fix this, you can either trigger the ArgoCD sync, or wait until the ArgoCD sync triggers the installer automatically to reinstall the packages. - After the environment upgrade, skill sync happens for all the skills and cannot be controlled for the selective ones.
- Prediction count doesn't increase in Automation Suite 2023.10 for all the ML Skills.