Introduction to Algorithm Preparation

To use a custom algorithm, use a framework built in ModelArts. ModelArts supports most mainstream AI engines. For details, see Built-in Training Engines. These built-in engines pre-load some extra Python packages, such as NumPy. You can also use the requirements.txt file in the code directory to install dependency packages. For details about how to create a training job using a preset image, see Using a Preset Image (Custom Script).

The subscribed algorithms and built-in frameworks can be used in most training scenarios. In certain scenarios, ModelArts allows you to create custom images to train models. Custom images can be used to train models in ModelArts only after they are uploaded to Software Repository for Container (SWR). For details, see Using a Custom Image to Train Models Customizing an image requires a deep understanding of containers. Use this method only if the subscribed algorithms and custom scripts cannot meet your requirements.

Offline learning is the most fundamental mode for model training. In this mode, all data required for training must be provided at a time, and optimizing the objective function stops when the training is complete. The advantage of this mode is that the trained models are stable, facilitating model verification and evaluation.

Incremental learning is a continuous learning process. Compared with offline learning, it does not need to store all training data at a time, which alleviates the problem of limited storage resources. In addition, it saves a large amount of compute power and time, and reduces economic costs in retraining.