Skip to main content

H2O AI Managed Cloud release notes

24.04.0 | April 01, 2024​

Upgraded components​

  • Upgraded AI App Store to v0.34.0
  • Upgraded H2O Discovery service to v3.2.0
  • Upgraded H2O MLOps to v0.64.0
  • Upgraded H2O AI Engine Manager to v0.7.0
  • Upgraded H2O Drive backend to v3.0.1
  • Upgraded H2O Driverless AI to v1.10.7
  • Upgraded H2O eScorer to v0.10.0
  • Upgraded H2O-3 to v3.46.0.1
  • Upgraded H2O Managed Cloud Admin Center to v0.12.0
  • Upgraded Enterprise H2OGPTe to v1.4.9
  • Upgraded AuthZ to v5.2.0
  • Upgraded Telemetry to v1.1.0

New features and improvements​

Platform​

  • Improved startup time for H2O Driverless AI instances.

AI App Store​

  • Added support for Python 3.12
  • Added the abilty to configure the CLI automatically using only the App Store URL and token.
info

For a full list of features, improvements, and bug fixes in AI App Store, see the AI App Store v0.31.2, v0.32.0, v0.33.0, & v0.34.0 release notes.

H2O Drive​

  • Added automatic H2O Drive discovery and authentication. Users of the H2O Drive Python client can now interact with H2O Drive immediately out-of-the-box, without having to configure their client with endpoints or retrieve and exchange access tokens themselves.
  • Speedier operations. Users should see their reads and writes complete up to 25% faster.
  • Fixed an issue wherein reading objects would sometimes incorrectly result in an error.
  • Enabled support for interacting with H2O Drive from outside the H2O Managed Cloud cluster. Contact your H2O Managed Cloud representative to check if this is right for your deployment.

H2O eScorer​

  • Added the ability to deploy models to remote H2O eScorer instances and monitor their stats.
  • The H2O eScorer Python Client now supports scoring Pandas dataframes.
  • The model registry can be filtered, sorted, and searched.
  • Added initial LLM support.
  • Upgraded Wave app to Wave SDK v1.0.2
  • AutoGen generates code with syntax highlighting.
  • Snowpark Container Services queries have been updated.
  • Snowpark Container Services can be suspended and resumed within the UI.
  • The eScorerAsyncClient class in Python Client was deprecated in favor of Client.
  • Scoring a model from Model Registry auto-fills the dropdown with model name.
  • Page layouts have been improved, and navigation panel groups are collapsed by default.
  • Added refresh button to refresh list of projects in H2O MLOps.
info

For a full list of features, improvements, and bug fixes in H2O eScorer, see the H2O eScorer v0.10.0 release notes.

H2O MLOps​

  • Added H2O Driverless AI v1.10.6.2 and v1.10.7 runtime support.
  • Extended the scoring API with a new endpoint /model/media-score to support uploading multiple media files and you can now score image and audio files using this endpoint on the H2O Hydrogen Torch runtime.
  • Added an Events page with pagination, search, and sorting.
  • Added the capability to delete experiments.
  • Added pagination, search, sorting, and filtering by Tag on the Experiments page.
  • The Create Deployment workflow now automatically populates K8s limits and requests with the suggested default settings.
  • The deployment state is now updated dynamically on the Deployments page.
  • Additional details about error deployment states are now displayed in the MLOps UI.
  • You can now update and delete tags. Note that tags can only be deleted if they are not associated with any entity.
  • MLflow Dynamic Runtime: Added support for Python 3.10.
  • Upgraded Rest scorer to Spring Boot 3 (1.2.0).
  • Added vLLM runtime support.
  • When creating a new deployment, added an option to disable monitoring for the deployment.
  • Added validation for experiment file uploading.
info

For a full list of features, improvements, and bug fixes in H2O MLOps, see the H2O MLOps v0.63.0 & v0.64.0 release notes.

H2O Driverless AI​

  • Added a new configuration that lets users use their own service account or select which service account to impersonate when connecting to Google BigQuery (GBQ)
  • Added a new configuration to control the experiment Leaderboard access globally for all users.
  • H2O DAI now reports why certain explainers are not enabled in the MLI explainer drop-down list. This is usually limited by the experiment problem type, e.g., image, multinomial, etc.
  • Users are now warned when Shapley values are approximated in MLI.
  • Added recipe support for the Feature Store data connector.
  • Added a new configuration that lets users change the timeout duration when importing data from Hive, HDFS, JDBC, and kdb+ connectors.
  • Ability to navigate to the linked project from the experiment details page.
  • Added support for Python 3.12 in h2oai_client.
  • MLI capabilities:
    • Added the 2-D Partial Dependence explainer.
    • Added the Friedman H-statistic explainer.
    • Added PDP percentile plot.
  • Added ICE curves in the Partial Dependence explainer at every decile of predicted probabilities. This gives an indication of local prediction behavior across the dataset.
  • The Python client now supports the generation of plots for more MLI explainers.
info

For a full list of features, improvements, and bug fixes in H2O Driverless AI, see the H2O Driverless AI v1.10.6.2 & v1.10.7 changelog.

H2O Managed Cloud Admin Center​

  • Added a new page for the information on daily cumulative hourly peak usage of AI units where the users can gain valuable insights into how their AI unit consumption fluctuates over time. This will help them to make informed decisions about their usage of AI units. The new AI unit hourly page includes two sections:

    • Cumulative summary: displays the daily cumulative hourly usage of AI units.
    • Component summary: displays the daily cumulative hourly peak usage of AI units per each application or product component installed in the environment.

    Each section features two plots:

    1. The daily cumulative hourly usage of AI units over the entire contract period of the HAMC environment.
    2. The daily cumulative hourly usage for a specific month. The users can select any month from the contract period.

  • Added a new menu card on the dashboard providing direct access to the new AI unit hourly page with a brief introduction to the AI unit hourly page.

info

For a full list of features, improvements, and bug fixes in H2O Managed Cloud Admin Center, see the H2O Managed Cloud Admin Center v0.12.0 release notes.

H2O-3​

  • Implemented a new write_checksum parameter that allows you to disable default Hadoop Parquet writer systematically writing a .crc checksum file for each written data file.
  • Implemented make_metrics with custom AUUC thresholds for UpliftDRF.
  • Implemented custom metric for AutoML.
  • Implemented custom metric for Stacked Ensemble.
  • Implemented MOJO support for UpliftDRF.
  • Implemented Python 3.10 and 3.11 support.
  • Implemented custom metric for Deep Learning.
  • Implemented custom metric functionalities and the ATE, ATT, and ATC metrics for UpliftDRF.
  • Implemented custom metric for leaderboard.
  • Implemented new AdaBoost algorithm for binary classification.
  • Implemented Shapley values support for ensemble models
  • Added anomaly score metric to be used as a sort_by metric when sorting grid model performances for Isolation Forest with grid search.
  • Added weak_learner_params parameter for AdaBoost.
  • Added weak_learner="deep_learning" option for AdaBoost.
  • Implemented scoring and scoring history for Extended Isolation Forest by adding score_each_iteration and score_tree_interval.
  • Added custom_metric as a hyperparameter for grid search.
  • Added custom metrics for XGBoost.
  • Implemented consistent mechanism that protects frames and their vecs from autodeletion.
  • Introduced a warning if web_ip is not specified that H2O Rest API is listening on all interfaces.
  • Introduced MLFlow flavors for working with H2O-3 MOJOs and POJOs instead of binary models.
  • Implemented machine learning interpretability support for UpliftDRF by allowing Uplift models to access partial dependences plots and variable importance.
info

For a full list of features, improvements, and bug fixes in H2O-3, see H2O-3 v3.42.0.3, v3.42.0.4, v3.44.0.1, v3.44.0.2, v3.44.0.3 & v3.46.0.1 changelog and the H2O-3 v.3.44.x release blog.

Enterprise H2OGPTe​

  • Added a get_default_collection function for Collection API keys.
  • Added support for external vector databases.
  • Added support for Milvus, Redis, and Qdrant as Vector databases.
  • The Collections UI now shows the number of Chat Sessions associated with each Collection.
  • Added support for the following:
  • Added FlagEmbedding BGE-M3, which lets the user choose the embedding model (English/Multilingual/Legacy) and is recommended for non-English or multi-lingual datasets.
  • Added H2O Danube, which is a mini LLM that can be fine-tuned to handle a range of NLP applications on mobile devices.
  • Added speaker diarization for audio. This helps to identify who said what within audio files.
  • Redesigned the Chat UI.
  • You can now connect to the Python API using the H2O AI Cloud API token.
  • Enterprise H2OGPTe now provides the option to import a website without crawling it.
  • Added initial speech recognition support using OS built-in speech recognition. Supported in Google Chrome and Safari browsers. This feature is still in its initial stages and only converts raw audio to text at the moment.
  • Added a catalog of Prompt Templates with support for many languages. You can now create your own custom Prompt Templates and select a Prompt Template to set the Prompts for a Collection or a specific Chat Session.
  • Admin users can now pull job information for all other users.

24.02.1 | February 18, 2024​

Upgraded components​

  • Upgraded H2O Feature Store to v1.2.0
  • Upgraded H2O eScorer to v0.9.0

New features​

Platform​

  • Migration to the H2O MLOps HELM chart
  • Fix H2O eScorer to pull the correct H2O Driverless AI license
  • Limit access to customer data from customer VPC only
  • Deny policy to H2O Drive S3 for all roles except drive-sa-role
  • New Terraform data module
  • Refactoring to environment wipe process
  • Access logging to tfstate S3
  • Removal of grafanaauth provider
  • Migration of all H2O Enterprise Steam instances to H2O AI Engine Manager

H2O Feature Store​

  • Introduce H2O Drive as a data source
info

For a full list of features, improvements, and bug fixes in H2O Feature Store, see the H2O Feature Store release notes.

H2O eScorer​

  • Support for Snowpark Container Services
  • New monitor charts
  • Ability to show statistics on training and scored tables
info

For a full list of features, improvements, and bug fixes in H2O eScorer, see the H2O eScorer release notes.

24.02.0 | February 04, 2024​

New components​

The following are new and optional components of H2O Managed Cloud:

  • H2O Eval Studio

Upgraded components​

  • Upgraded AI App Store to v0.31.1
  • Upgraded AuthZ to v4.0.2
  • Upgraded Enterprise h2oGPTe to v1.3.4
  • Upgraded H2O AI Notebooks to v0.3.0
  • Upgraded Telemetry to v0.30.3
  • Upgraded H2O Drive to v0.8.1
  • Upgraded Managed Cloud Admin Center to v0.11.0

New features​

Platform​

  • Block creation of new H2O Driverless AI and H2O-3 instances in Enterprise Steam
    • Existing instances can still be started
    • New instances should be started using H2O AI Engine Manager (AIEM)
    • All existing instances in Enterprise Steam will be migrated to AIEM on the 18.03.2024
  • Added a new Beamer side panel for Managed Cloud news, announcements, and latest release notes.

AI App Store​

  • Fixed a UI style issue with Beamer
info

For a full list of features, improvements, and bug fixes in H2O AI App Store, see the H2O AI App Store release notes.

Managed Cloud Admin Center​

  • Upgraded PyJWT version.
  • Added a new page for information about H2O Engine Manager (AIEM), which provides useful AIEM insights of the Managed Cloud environment.
  • Added a new menu card on the dashboard for the new AIEM insights page.
info

For a full list of features, improvements, and bug fixes in H2O Managed Cloud Admin Center, see the H2O Admin Center release notes.

Enterprise h2oGPTe​

  • Many significant changes and new features were added to this application since the previous version of H2O Managed Cloud.
info

For more information about all the H2O Enterprise h2oGPTe capabilities, see the h2oGPTe documentation.

H2O Drive​

  • Fixed a bug with the application crashing when no databases are found.
  • Added debug logging when connecting to Snowflake.
info

For a full list of features, improvements, and bug fixes in H2O Drive, see the H2O Drive release notes.

v23.12.1 | December 18, 2023​

  • Upgraded H2O AI Engine Manager to v0.6.0
  • Upgraded AI App Store to v0.30.1
  • Upgraded AuthZ to v2.1.0
  • Upgraded H2O Enterprise GPT to v1.1.7
  • Upgraded H2O Feature Store to v1.1.2
  • Upgraded h2oGPT to v0.1.0-180
  • Upgraded H2O ML-API to v0.6.2
  • Upgraded Managed Cloud Admin Center to v0.10.1
  • Upgraded H2O Label Genie to v0.4.0
  • Upgraded H2O Model Validation to v0.17.0

v23.11.0 | November 05, 2023​

  • Added H2O Notebook Lab v0.2.3
  • Added H2O LLM Studio v1.0.1
  • Added H2O LLM Data Studio
  • Added H2O Enterprise GPT to v0.9.2
  • Added h2oGPT v0.1.0-128
  • Upgraded H2O Driverless AI to v1.10.6.1
  • Upgraded Managed Cloud Admin Center to 0.10.0
  • Upgraded H2O Engine Manager to v0.5.4
  • Upgraded AI App Store to v0.29.1
  • Upgraded H2O Document AI to v0.7.0
  • Upgraded H2O Drive to v0.8.0
  • Upgraded H2O MLOps to v0.62.1
  • Upgraded H2O ML-API to v0.6.1
  • Upgraded AuthZ to v3.2.2
  • Upgraded Cloud Discovery to v3.0.2
  • Deprecated PY3.7 GPU runtime
Important

Please note that Enterprise Steam will be deprecated on the 11th of December 2023 and all H2O Driverless AI instances will be migrated to H2O AI Engine Manager.

v23.09.0 | September 10, 2023​

  • Added H2O Driverless AI v1.10.5.1
  • Added H2O-3 v3.42.0.2
  • Upgraded H2O AI AppStore to v0.28.2
  • Upgraded Managed Cloud Admin Center to v0.9.0
  • Upgraded Cloud Discovery v2.2.2
  • Upgraded Telemetry to v0.28.0
  • Upgraded AuthZ to v3.1.2
  • Upgraded H2O MLOps to v0.62.0
  • Upgraded H2O AI Engine Manager to v0.5.1
  • Upgraded Enterprise Steam to v1.9.6
  • Upgraded H2O Document AI to FE v0.8.3, Proxy v0.2.21, Scorer v0.2.22, and Viewer v0.8.4
  • Upgraded H2O ML-API to v0.5.3
  • Upgraded H2O eScorer to v0.6.1
  • Upgraded H2O Feature Store to v0.19.3

v23.08.1 | August 06, 2023​

Improvements​

  • Upgraded H2O Feature Store to v0.19.1
  • Upgraded Managed Cloud Admin Center to v0.8.0
  • Upgraded H2O Model Validation application to v0.16.0
  • Upgraded H2O AutoDoc to v0.8.5
  • Upgraded H2O Model Analyzer to v0.7.3

v23.06.1 | June 25, 2023​

New Features​

  • New AWS region; us-west-2 (Oregon)
  • Added support for BYOI (Bring Your Own Identity) using Okta

Improvements​

  • Added attributes “Supported”, “Preview”, and “Deprecated” to applications
  • Upgraded AI AppStore to v0.25.2
  • Upgraded H2O Feature Store v0.18.0
  • Upgraded H2O MLOps to v0.61.1
  • Upgraded H2O Drive to v0.7.4
  • Upgraded H2O Subsalt to v1.5.9
  • Upgraded H2O Managed Cloud Admin Center to v0.7.4
  • Upgraded AutoInsights to v0.8.1

v23.05.1 | May 21, 2023​

New Features​

  • Added H2O Document AI Viewer
  • Added H2O Admin Analytics application v3.2.0
  • Added H2O-3 v3.40.0.4
  • Added H2O Driverless AI v1.10.5
  • Added H2O Discovery Service

Improvements​

  • Increased the H2O Driverless AI timeout to 72 hours
  • Upgraded Enterprise Steam to v1.9.4
  • Upgraded AI App Store to v0.25.1
  • Upgraded H2O MLOps to v0.61.0
  • Upgraded H2O Document AI to v0.6.0
  • Upgraded H2O Managed Cloud Admin Center to v0.7.0
  • Upgraded H2O Drive to v0.7.3
  • Upgraded H2O eScorer to v0.5.0

v23.04.2 | April 16, 2023​

  • Upgraded H2O Hydrogen Torch to v1.3.0
  • Upgraded H2O Label Genie to v0.3.0
  • Upgraded H2O Auto Insights to v0.7.4

v23.04.1 | April 2, 2023​

Improvements​

  • Upgraded AI AppStore to v0.23.0
  • Upgraded H2O Feature Store to v0.14.4
  • Upgraded Enterprise Steam to v1.9.0
  • Upgraded H2O eScorer to v0.4.0
  • Upgraded H2O Managed Cloud Admin Center to v0.6.1

v23.03.1 | March 05, 2023​

Improvements​

  • Added H2O Driverless AI v1.10.4.3
  • Upgraded H2O MLOps to v0.59.1
  • Upgraded AI Engine Manager to v0.2.4
  • Upgraded H2O Drive to v0.7.2
  • Upgraded Managed Cloud Admin Center to v0.6.0
  • Removed operational overhead of adding domains to the include-list in the Managed Cloud Admin Center

Security improvements​

  • Automatically updated exclude-list of malicious domains for outbound traffic

v23.02.1 | February 12, 2023​

Improvements​

  • Added H2O Driverless AI v1.10.4.1
  • Added Jupyter Notebook XL
  • Upgraded AI AppStore to v0.22.0
  • Upgraded H2O MLOps to v0.59.0
  • Upgraded H2O Feature Store to v0.13.0
  • Upgraded H2O Document AI to v0.5.0
  • Upgraded Enterprise Steam to v1.9.0
  • Upgraded H2O Model Analyzer to v0.7.0
  • Upgraded H2O AutoInsights to v0.7.3
  • Upgraded H2O Driverless AI Quick Start to v0.2.5
  • Updated H2O Managed Cloud Admin Center v0.5.3

v22.12.1 | December 18, 2022​

Improvements​

  • Added tags to distinguish between Supported, Open Source, and Demo AppStore applications
  • Updated H2O eScorer application
  • Updated Subsalt application
  • Updated H2O Drive application
  • Updated Managed Cloud Admin Center application
  • Updated H2O Model Validation application

Security improvements

  • Added SSL encryption for internal database traffic
  • Updated threat intelligence source to include over 100,000 indicators of compromise

Fixes​

Updated Steam profile naming to align with CPU and GPU profiles

v22.11.1 | November 06, 2022​

New Features​

  • New Managed Cloud region in Mumbai

Improvements​

  • Added DAI v1.10.4
  • Added MLOps application
  • Updated MLOps to v0.57.3
  • Updated H2O Drive application
  • Updated Managed Cloud Admin Center application
  • Updated Model Monitoring application

v22.10.1 | October 02, 2022​

New features​

  • New Managed Cloud region in Singapore
  • Configure AI unit usage email alerting using Managed Cloud Admin Center
  • BYOI (bring your own identity) using Azure Active Directory
  • Label Genie application
  • Model Analyzer application (available for GPU enabled environments)

Improvements​

  • DAI launch time reduced to ~3 minutes
  • Added H2O-3 3.36.1.3
  • Updated AppStore to 0.21.0
  • Updated FeatureStore to 0.9.0
  • Updated Steam to 1.18.4
  • Updated Document.ai to 0.4.0
  • Updated Model Monitoring application
  • Updated Model Validation application
  • Added HydrogenTorch 1.2.0

v22.08.1 | August 23, 2022​

New features​

  • H2O eScorer
  • DAI Quickstart application
  • DAI Rule Explainer application

Improvements​

  • Updated Steam to 1.8.13
  • Updated Managed Cloud Admin Center
  • Updated H2O Drive
  • Updated Snowflake Integration application

Fixes​

  • Fixed Steam public URL in AppStore

v22.07.1 | July 10, 2022​

New features​

  • H2O Document AI
  • H2O Feature Store
  • Added H2O AI Cloud home page
  • H2O AI Notebooks application
  • Added Google Authenticator and FIDO 2 as additional MFA methods
  • H2O Drive import manager, multiple file upload, and support for SQL queries
  • Managed cloud admin center now includes a new AI unit consumption view

Improvements​

  • Updated AppStore to 0.19.1
  • Updated MLOps to 0.57.0
  • Updated Steam to 1.8.12
  • Added DAI 1.10.3 and 1.10.3.1
  • Added H2O-3 3.36.0.3
  • Updated Managed Cloud Admin Center
  • Updated H2O wave tour application
Feedback