Introduction
ModelArts supports multiple types of training jobs during the entire AI development process. Select a creation mode based on the algorithm source.
Algorithm Sources of Training Jobs
- Algorithm Management
In Algorithm Management, the algorithms created by users and subscribed algorithms in AI Gallery are managed. You can use algorithms in Algorithm Management to quickly create training jobs and build models. For details about the operation guide, see Using Existing Algorithms to Train Models. For details about algorithm management, see Overview.
- Built-in
If you do not know how to develop algorithms, you can use the built-in algorithms of ModelArts. After simple parameters adjustment, you can create a training job and build a model. For details about the operation guide, see Using Built-in Algorithms to Train Models. For details about the existing built-in algorithms and running parameters, see Introduction to Built-in Algorithms.
- Frequently-used
If you have used some frequently-used frameworks to develop algorithms locally, you can select a frequently-used framework and create a training job to build a model. For details, see Using Frequently-used Frameworks to Train Models.
- Custom
If the framework used for algorithm development is not a frequently-used framework, you can build an algorithm into a custom image and use the custom image to create a training job. For details about the operation guide to create a training job, see Using Custom Images to Train Models. For details about the specifications and description of custom images, see Specifications for Custom Images Used for Training Jobs.
Last Article: Creating a Training Job
Next Article: Using Existing Algorithms to Train Models
Did this article solve your problem?
Thank you for your score!Your feedback would help us improve the website.