Help Center> ModelArts> FAQs> General Issues> Which AI Frameworks Does ModelArts Support?

View PDF

Which AI Frameworks Does ModelArts Support?

The AI frameworks and versions supported by ModelArts vary slightly based on the development environment, training jobs, and model inference (model management and deployment). The following describes the AI frameworks supported by each module.

Development Environment

The AI frameworks and versions supported by DevEnviron notebooks vary based on Python versions.

**Table 1** AI engines supported by old development environments
Work Environment	Built-in AI Engine and Version	Supported Chip
Multi-Engine 1.0 (Python3, Recommended)	MXNet-1.2.1	CPU/GPU
	PySpark-2.3.2	CPU
	PyTorch-1.0.0	GPU
	TensorFlow-1.13.1	CPU/GPU
	TensorFlow-1.8	CPU/GPU
	XGBoost-Sklearn	CPU
Multi-Engine 1.0 (Python2)	Caffe-1.0.0	CPU/GPU
	MXNet-1.2.1	CPU/GPU
	PySpark-2.3.2	CPU
	PyTorch1.0.0	GPU
	TensorFlow-1.13.1	CPU/GPU
	TensorFlow-1.8	CPU/GPU
	XGBoost-Sklearn	CPU
Multi-Engine 2.0 (Python3)	PyTorch-1.4.0	GPU
	R-3.6.1	CPU/GPU
	TensorFlow-2.1.0	CPU/GPU
Ascend-Powered-Engine 1.0 (Python3)	MindSpore-1.1.1	Ascend 910
Ascend-Powered-Engine 1.0 (Python3)	TensorFlow-1.15.0	Ascend 910

Training Jobs

ModelArts provides training jobs of both old and new versions. The difference between the two versions lies in built-in training engines. You are suggested to choose the new version to use new built-in training engines. The engines in the old training job version can also be used in the new version.

**Table 2** AI engines supported by the original training version
Work Environment	Supported Chip	System Architecture	System Version	AI Engine and Version	Supported CUDA or Ascend Version
TensorFlow	CPU/GPU	x86_64	Ubuntu16.04	TF-1.8.0-python3.6	cuda9.0
				TF-1.8.0-python2.7	cuda9.0
				TF-1.13.1-python3.6	cuda10.0
				TF-1.13.1-python2.7	cuda10.0
				TF-2.1.0-python3.6	cuda10.1
MXNet	CPU/GPU	x86_64	Ubuntu16.04	MXNet-1.2.1-python3.6	cuda9.0
MXNet	CPU/GPU	x86_64	Ubuntu16.04	MXNet-1.2.1-python2.7	cuda9.0
Caffe	CPU/GPU	x86_64	Ubuntu16.04	Caffe-1.0.0-python2.7	cuda8.0
Spark_MLlib	CPU	x86_64	Ubuntu16.04	Spark-2.3.2-python2.7	-
Spark_MLlib	CPU	x86_64	Ubuntu16.04	Spark-2.3.2-python3.6	-
Ray	CPU/GPU	x86_64	Ubuntu16.04	RAY-0.7.4-python3.6	cuda10.0
XGBoost-Sklearn	CPU	x86_64	Ubuntu16.04	Scikit_Learn-0.18.1-python2.7	-
XGBoost-Sklearn	CPU	x86_64	Ubuntu16.04	Scikit_Learn-0.18.1-python3.6	-
PyTorch	CPU/GPU	x86_64	Ubuntu16.04	PyTorch-1.0.0-python2.7	cuda9.0
				PyTorch-1.0.0-python3.6	cuda9.0
				PyTorch-1.3.0-python2.7	cuda10.0
				PyTorch-1.3.0-python3.6	cuda10.0
				PyTorch-1.4.0-python3.6	cuda10.1
Ascend-Powered-Engine	Ascend 910	aarch64	Euler2.8	Mindspore-1.1.1-python3.7-aarch64	21.0.0
Ascend-Powered-Engine	Ascend 910	aarch64	Euler2.8	TF-1.15-python3.7-aarch64	21.0.0
MindSpore-GPU	CPU/GPU	x86_64	Ubuntu18.04	MindSpore-1.1.0-python3.7	cuda10.1

The built-in training engines in the new version are named in the following format:

<Training engine name_version>-[cpu | <cuda_version | cann_version>] --<py_version><OS name_version>-< x86_64 | aarch64>

**Table 3** AI engines supported by the new training version
Work Environment	Supported Chip	System Architecture	System Version	AI Engine and Version	Supported CUDA or Ascend Version
TensorFlow	CPU/GPU	x86_64	Ubuntu18.04	tensorflow_2.1.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64	cuda10.1
PyTorch	CPU/GPU	x86_64	Ubuntu18.04	pytorch_1.8.0-cuda_10.2-py_3.7-ubuntu_18.04-x86_64	cuda10.2
Ascend-Powered-Engine	Ascend 910	aarch64	Euler2.8	mindspore_1.3.0-cann_5.0.2-py_3.7-euler_2.8.3-aarch64	cann_5.0.2
Ascend-Powered-Engine	Ascend 910	aarch64	Euler2.8	tensorflow_1.15-cann_5.0.2-py_3.7-euler_2.8.3-aarch64	cann_5.0.2
MPI	CPU/GPU	x86_64	Ubuntu18.04	mindspore_1.3.0-cuda_10.1-py_3.7-ubuntu_1804-x86_64	cuda_10.1
Horovod	GPU	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	GPU	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
KungFu	CPU/GPU	x86_64	Ubuntu18.04	KF-0.2.2-TF-1.13.1-python3.6	-

Ascend-Powered-Engine is only available in the CN North-Beijing4 region.

Model Inference

For imported models and model inference is completed on ModelArts, supported engines and their runtime are as follows:

**Table 4** Supported engines and their runtime
Engine	Runtime	Remarks
TensorFlow	python3.6 python2.7 tf1.13-python2.7-gpu tf1.13-python2.7-cpu tf1.13-python3.6-gpu tf1.13-python3.6-cpu tf1.13-python3.7-cpu tf1.13-python3.7-gpu tf2.1-python3.7	TensorFlow 1.8.0 is used in python2.7 and python3.6. python3.6, python2.7, and tf2.1-python3.7 indicate that the model can run on both CPUs and GPUs. For other runtime values, if the suffix contains cpu or gpu, the model can run only on CPUs or GPUs. The default runtime is python2.7.
MXNet	python3.7 python3.6 python2.7	MXNet 1.2.1 is used in python2.7, python3.6, and python3.7. python2.7, python3.6, and python3.7 indicate that the model can run on both CPUs and GPUs. The default runtime is python2.7.
Caffe	python2.7 python3.6 python3.7 python2.7-gpu python3.6-gpu python3.7-gpu python2.7-cpu python3.6-cpu python3.7-cpu	Caffe 1.0.0 is used in python2.7, python3.6, python3.7, python2.7-gpu, python3.6-gpu, python3.7-gpu, python2.7-cpu, python3.7-cpu, and python3.6-cpu. python 2.7, python3.7, and python 3.6 can only be used to run models applicable to CPU. For other runtime values, if the suffix contains cpu or gpu, the model can run only on CPUs or GPUs. You are advised to use the runtime of python2.7-gpu, python3.6-gpu, python3.7-gpu, python2.7-cpu, python3.7-cpu, and python3.6-cpu. The default runtime is python2.7.
Spark_MLlib	python2.7 python3.6	Spark_MLlib 2.3.2 is used in python2.7 and python3.6. The default runtime is python2.7. python 2.7 and python 3.6 can only be used to run models applicable to CPU.
Scikit_Learn	python2.7 python3.6	Scikit_Learn 0.18.1 is used in python2.7 and python3.6. The default runtime is python2.7. python 2.7 and python 3.6 can only be used to run models applicable to CPU.
XGBoost	python2.7 python3.6	XGBoost 0.80 is used in python2.7 and python3.6. The default runtime is python2.7. python 2.7 and python 3.6 can only be used to run models applicable to CPU.
PyTorch	python2.7 python3.6 python3.7 pytorch1.4-python3.7	PyTorch 1.0 is used in python2.7, python3.7, and python3.6. python2.7, python3.6, python3.7, and pytorch1.4-python3.7 indicate that the model can run on both CPUs and GPUs. The default runtime is python2.7.

Parent topic: General Issues

Did this article solve your problem?

Thank you for your score！Your feedback would help us improve the website.

Products

Compute

Application

Dedicated Cloud

Storage

Management & Deployment

Migration

Network

Enterprise Intelligence

Video

Database

Edge Cloud Services

DevCloud

Security

Cloud Communications

Internet of Things

Solutions

Industry-Specific Solutions

General-Purpose Solutions

Security

DevOps

Enterprise Intelligence

Essential Platform

Big Data

Visual Cognition

Speech and Semantics

Support

Help Center

Customer Services

Developers

Console

语言 - Language

中国站 - 简体中文

中国站 - English

International - 简体中文

International - English

Help Center

Which AI Frameworks Does ModelArts Support?

Development Environment

Training Jobs

Model Inference