Change History

2023-03-30

Deleted "DevEnviron (Old Version)".

Deleted section "Creating AI Applications" in "Custom Image".

2023-03-01

Moved the content in "Audit Logs" to Audit Logs in "Resource Management".

Moved the content in "Permissions Management" to Permissions Management in "Best Practices".

Removed "Operation Guide".

2022-12-30

Moved the content in "AI Application Management", "Deploying a Service", "Model Package Specifications", "Model Version", "Examples of Custom Scripts", and "Monitoring" to Inference Deployment.

2022-05-19

Added hitless rolling upgrade for inference services.

2022-01-26

Removed "Hard Example Filtering" in "Deploying a Service".

2022-01-04

Discontinued one-click model deployment in "Data Management".

2021-12-20

Discontinued built-in algorithms in the training management of the old version.

Canceled "Evaluating Models" because model evaluation has been changed to a Closed Beta Test (CBT) function.

2021-12-15

Changed "Model Management" to "AI Application Management".

2021-10-21

Added "Differences Between the New and Old Versions of Training".

2021-06-26

Added "Training Management (New Version)".

2020-01-14

• Optimized the description for the following section:

  Creating and Uploading a Custom Image

• Added the machine learning inference code example and configuration file description in model package specifications.
• Optimized the PyTorch code example in the customized script.

2020-11-10

• Optimized the display of the development environment when a notebook is created.

2020-09-21

• Added the application authentication function for real-time services.

2020-08-06

• Updated the description of the notebook function, optimized the document structure, and added common JupyterLab operations.
• Added the dataset of the table type, which supports the DWS data source input.
• Added guidance on how to integrate real-time service APIs.

2020-07-06

Optimized auto search jobs and added one sample.

Added the support for the GitHub code library in the development environment.

Added the support for the PyTorch 1.4.0 engine in development environments and model import.

Added the sample code of the TensorFlow 2.X custom script.

2020-06-08

Added table dataset and CSV file import and release.

Enhanced DevEnviron by introducing JupyterLab and TensorFlow 2.1, integrated multiple basic capabilities of ModelArts, and improved user experience.

Optimized auto search job: Added hyperparameter search with fix_norm, data augmentation with adv_aug, and multiple NAS methods such as BETANAS to achieve the optimal precision in mobile setting on ImageNet. Multisearch is available using at least five lines of code.

Added TensorFlow 2.1. When creating a notebook instance or using a frequently-used framework to create a training job, TensorFlow 2.1 is ready to use.

• Using Frequently-used Frameworks to Train Models

Added the Python 3.7 runtime environment to Model Management.

2020-03-24

• Optimized the description of training jobs based on different algorithm sources.
  • Using Frequently-used Frameworks to Train Models
  • Using Custom Images to Train Models
• Added support for team labeling for datasets of the text triplet, text classification, and named entity types, and updated the description of the data labeling function.
• Updated all screenshots in this document because ModelArts UI is upgraded.

2020-01-07

• Hid the old data management module.

2019-12-03

• Added four sections for the model import function based on scenarios.
• Supported team labeling task management for datasets that support team labeling.
• Added the text triplet dataset type.
• Optimized the sections about data management.
• Added the function of managing auto search jobs.

2019-10-17

• Optimized and added functions for data management based on software changes. Updated descriptions in all related sections and added the following content:
• Added sample code of custom scripts (including frequently-used engines).
• Added monitoring description.

2019-09-30

• Enriched and optimized the description of model templates.
• Optimized the description of model package specifications and provided more model package examples.

2019-08-20

Added data management and model templates.

2019-06-13

This is the first official release.