Creation Methods

AI development and optimization require frequent iterations and debugging. Modifications in datasets, training code, or parameters affect the quality of models. If the metadata of the development process cannot be centrally managed, the optimal model may fail to be reproduced.

With ModelArts, you can create models using meta models from training jobs, OBS, or container images, and centrally manage all iterated and debugged models.

Constraints

After deploying a model in an ExeML project, it is automatically added to the model list. ExeML-generated models can only be deployed, not downloaded.
All users can create models and manage model versions at no cost.

Meta Model Sources

Importing a Meta Model from a Training Job: Create a training job in ModelArts to train a model. After obtaining a desired model, use it to create a model for service deployment.
Importing a Meta Model from OBS: If you use a mainstream framework to develop and train a model locally, you can upload the model to an OBS bucket based on the model package specifications, import the model from OBS to ModelArts, and use the model for service deployment.
Importing a Meta Model from a Container Image: If an AI engine is not supported by ModelArts, you can use it to build a model, import the model to ModelArts as a custom image, and use the image to create a model for service deployment.

Supported AI Engines for ModelArts Inference

If you import a model from OBS to ModelArts, the following AI engines and versions are supported.

A runtime environment of a unified image is named in the following format: <AI engine and version> - <Hardware and version: CPU, CUDA, or CANN> - <Python version> - <OS version> - <CPU architecture>
Each preset AI engine has its default model start command. Do not modify it unless necessary.

**Table 1** Supported AI engines, their runtime environments, and default start commands
Engine	Runtime Environment	Note
TensorFlow	python3.6 python2.7 (unavailable soon) tf1.13-python3.6-gpu tf1.13-python3.6-cpu tf1.13-python3.7-cpu tf1.13-python3.7-gpu tf2.1-python3.7 (unavailable soon) tensorflow_2.1.0-cuda_10.1-py_3.7-ubuntu_18.04-x86_64 (recommended)	TensorFlow 1.8.0 is used in python2.7 and python3.6. The model can run on both CPUs and GPUs when using python3.6, python2.7, or tf2.1-python3.7. If the runtime environment has a suffix of cpu or gpu, the model can only run on CPUs or GPUs respectively. The default runtime environment is python2.7. Default start command: sh /home/mind/run.sh
Spark_MLlib	python2.7 (unavailable soon) python3.6 (unavailable soon)	Spark_MLlib 2.3.2 is used in python2.7 and python3.6. The default runtime environment is python2.7. python2.7 and python3.6 can only be used to run models on CPUs. Default start command: bash /home/work/predict/bin/run.sh
Scikit_Learn	python2.7 (unavailable soon) python3.6 (unavailable soon)	Scikit_Learn 0.18.1 is used in python2.7 and python3.6. The default runtime environment is python2.7. python2.7 and python3.6 can only be used to run models on CPUs. Default start command: bash /home/work/predict/bin/run.sh
XGBoost	python2.7 (unavailable soon) python3.6 (unavailable soon)	XGBoost 0.80 is used in python2.7 and python3.6. The default runtime environment is python2.7. python2.7 and python3.6 can only be used to run models on CPUs. Default start command: bash /home/work/predict/bin/run.sh
PyTorch	python2.7 (unavailable soon) python3.6 python3.7 pytorch1.4-python3.7 pytorch1.5-python3.7 (unavailable soon) pytorch_1.8.0-cuda_10.2-py_3.7-ubuntu_18.04-x86_64 (recommended)	PyTorch 1.0 is used in python2.7, python3.6, and python3.7. The model can run on both CPUs and GPUs when using python2.7, python3.6, python3.7, pytorch1.4-python3.7, or pytorch1.5-python3.7. The default runtime environment is python2.7. Default start command: sh /home/mind/run.sh
MindSpore	aarch64 (recommended)	aarch64 can only be used to run models on Snt3 chips. Default start command: sh /home/mind/run.sh