Overview

If the subscribed algorithms cannot meet your requirements or you want to migrate local algorithms to ModelArts for training, use the ModelArts preset images to create algorithms. This method is also called using a preset image.

This section describes how to use a preset image to create an algorithm.

For details about ModelArts built-in engines and models, see Built-in Training Engines.
To migrate local algorithms to ModelArts, perform code adaptation. For details, see Developing a Custom Script.
For details about how to use a preset image to create an algorithm on the ModelArts console, see Creating an Algorithm.

Built-in Training Engines

The following table lists the training engines and their versions supported by ModelArts.

Supported AI engines vary depending on regions.

**Table 1** AI engines supported by training jobs
Runtime Environment	System Architecture	System Version	AI Engine and Version	Supported CUDA or Ascend Version
TensorFlow	x86_64	Ubuntu18.04	tensorflow_2.1.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64	cuda10.1
PyTorch	x86_64	Ubuntu18.04	pytorch_1.8.0-cuda_10.2-py_3.7-ubuntu_18.04-x86_64	cuda10.2
MPI	x86_64	Ubuntu18.04	mindspore_1.3.0-cuda_10.1-py_3.7-ubuntu_1804-x86_64	cuda_10.1
Horovod	x86_64	ubuntu_18.04	horovod_0.20.0-tensorflow_2.1.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64	cuda_10.1
Horovod	x86_64	ubuntu_18.04	horovod_0.22.1-pytorch_1.8.0-cuda_10.2-py_3.7-ubuntu_18.04-x86_64	cuda_10.2