Help Center> ModelArts> User Guide (Senior AI Engineers)> DevEnviron (Notebook)> Introduction to Notebook
View PDF

Introduction to Notebook

ModelArts integrates the open source Jupyter Notebook and JupyterLab to provide you with online interactive development and debugging environments. You can use the Notebook on the ModelArts management console to compile and debug code and train models based on the code, without concerning installation and configurations.

  • Jupyter Notebook is an interactive notebook. For details about how to perform operations on Jupyter Notebook, see Jupyter Notebook Documentation.
  • JupyterLab is an interactive development environment. It is a next-generation product of Jupyter Notebook. JupyterLab enables you to compile notebooks, operate terminals, edit MarkDown text, open interaction modes, and view CSV files and images. For details about how to perform operations on JupyterLab, see JupyterLab Documentation.

Supported AI Engines

All supported AI engines can be used in the same notebook instance. Different engines can be switched quickly and conveniently, and run in independent development environments. Conda-python2 is a basic Python 2.7 environment without the AI engine. Conda-python3 is a basic Python 3.6 environment without the AI engine.

  • ModelArts notebook instances support multiple engines. That is, a notebook instance can use all supported engines. Different engines can be switched quickly and conveniently.
  • When creating a notebook instance, select a work environment that contains multiple engines, including Python2 and Python3. Multi-Engine 1.0 (Python3) is recommended. Select an applicable work environment. For details, see Table 1.
Table 1 AI engines

Work Environment

Built-in AI Engine and Version

Matching Chip

Multi-Engine 1.0 (Python 3, 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-1.0.0

Ascend 910

TensorFlow-1.15.0

Ascend 910

Constraints

  • For security purposes, the root permission is not granted to the notebook instances integrated in ModelArts. You can use the non-privileged user jovyan or ma-user (using Multi-Engine) to perform operations. Therefore, you cannot use apt-get to install the OS software.
  • Notebook instances support only standalone training under the current AI engine framework. If you need to use distributed training, you are advised to use ModelArts training jobs and specify multiple nodes in the resource pool.
  • ModelArts DevEnviron does not support apt-get. Instead, you can use Custom Image Overview.
  • Notebook instances do not support GUI-related libraries, such as PyQt.
  • Notebook instances created using Ascend specifications cannot be attached to EVS disks.
  • Notebook instances cannot be connected to DWS and database services.
  • Notebook instances cannot directly read files in OBS. You need to download the files to the local host. To access data in OBS, you are advised to use MoXing or SDK for interaction.
  • DevEnviron does not support TensorBoard. You are advised to use the visualization job function under Training Jobs.
  • After a notebook instance is created, you cannot modify its specifications. For example, you cannot change the CPU specifications to GPU specifications or change the work environment. Therefore, you are advised to select the specifications required by the service when creating a notebook instance, or save your code and data to OBS in a timely manner during development so that you can quickly upload the code and data to a new notebook instance.
  • If the code output is still displayed after you close the page and open it again, use Terminal.
  • After upgrading AI frameworks such as PyTorch, check whether they are compatible with the current CUDA version. The CUDA versions preset in Multi-Engine 2.0 are different from those preset in Multi-Engine 1.0 . For details about how to switch CUDA versions, see Switching the CUDA Version on the Terminal Page of a GPU-based Notebook Instance.
Parent topic: DevEnviron (Notebook)