Data Science Workshop (DSW) of Machine Learning Platform for AI (PAI) is an integrated development environment (IDE) in the cloud. DSW provides interactive development environments for developers of different levels. This topic describes the features, instance types, and supported zones of DSW editions. The editions are Individual Edition, GPU On-sale Edition, and Explorer Edition.

DSW integrates open source JupyterLab and provides plug-ins for customized development. You can directly launch Notebook to write, debug, and run Python code without O&M configurations. DSW also provides a variety of computing resources and supports heterogeneous data sources. DSW offers all-in-one machine learning services. You can use EASCMD to deploy models trained in DSW as RESTful APIs to provide online services.

Features

  • Supports real-time resource monitoring. Monitors CPU or GPU utilization during algorithm development.
  • Supports a variety of data sources, such as MaxCompute, Object Storage Service (OSS), and Apsara File Storage NAS.
  • Allows you to write and execute SQL statements.
  • Supports multiple resource types, including a variety of CPUs and GPU models.
  • Allows you to switch among different types of resources, which reduces resource usage costs.
  • Provides built-in big data development packages and algorithm libraries, and allows you to install third-party libraries.

Editions

DSW provides the following editions: Individual Edition, GPU On-sale Edition, and Explorer Edition. The following table describes the differences among the editions.
Feature Individual Edition GPU On-sale Edition Explorer Edition
GPUs Supported. Supported. Supported.
Pay-as-you-go billing method Supported. Not supported. Free of charge.
Subscription billing method Not supported. Supported. Free of charge.
Memory and number of CPU cores No upper limit. You can specify the memory and CPU cores as needed. No upper limit. You can specify the memory and CPU cores as needed. 2 vCPUs + 4 GB of memory
Instance storage No upper limit. You can specify storage space as needed. No upper limit. You can specify storage space as needed. 5 GB
Network access No limit. DSW instances of this edition cannot access the Internet. No limit on CPU-accelerated DSW instances. However, GPU-accelerated DSW instances cannot access the Internet.
Root permissions Supported. Not supported. Not supported.
Deploy a runtime from a specified image Supported. Not supported. Not supported.

Individual Edition

DSW Individual Edition is upgraded from DSW 2.0. This edition is developed based on cloud-native technologies of Alibaba Cloud, such as Docker and Kubernetes. It provides open and AI-assisted development environments for you to train models with high elasticity. The following section describes the features, instance types, and supported zones and images of this edition.
  • Features
    • Reduces the time that is required to create a DSW instance by 65%, compared with DSW 2.0. Does not charge additional fees for elastic IP addresses (EIPs) and Server Load Balancer (SLB) instances.
    • Allows you to start and stop DSW instances on demand, save images with one click, and restore development environments.
    • Provides integrated development environments:
      • Provides built-in big data development packages and algorithm packages, and grants sudo permissions to you for installing third-party libraries.
      • Provides built-in JupyterLab plug-ins to improve development efficiency, such as Git and TensorBoard.
      • Provides official images that support different versions of mainstream computing frameworks, such as TensorFlow and PyTorch.
      • Provides built-in WebIDE that allows you to install all plug-ins.
    • Supports basic features of PAI, including EasyVision, AutoML, TAO, and CommonIO. EasyVision is a computer vision algorithm tool. AutoML can help you automatically tune parameters. TAO is used to optimize compilation. CommonIO allows you to read data from MaxCompute tables.
    • Supports root permissions.
  • Instance types and zones
    The following table describes the instance types and supported zones of CPU-accelerated DSW instances.
    Instance type vCPUs Memory (GiB) Bandwidth (Gbit/s) System disk (GB) Region
    ecs.c6.large 2 4 1 128
    • China (Beijing)
    • China (Shanghai)
    • China (Hangzhou)
    • China (Shenzhen)
    ecs.g6.large 2 8 1 128
    ecs.g6.xlarge 4 16 1,5 256
    ecs.g6.2xlarge 8 32 2.5 500
    ecs.g6.4xlarge 16 64 5 500
    ecs.g6.8xlarge 32 128 10 500
    The following table describes the instance types and supported zones of GPU-accelerated DSW instances.
    Instance type vCPUs Memory (GiB) GPU Bandwidth (Gbit/s) System disk (GB) Region
    ecs.gn6e-c12g1.12xlarge 48 368 4*NVIDIA V100 16 500
    • China (Beijing)
    • China (Shanghai)
    • China (Hangzhou)
    • China (Shenzhen)
    'ecs.gn5-c4g1.xlarge 4 30 1*NVIDIA P100 3 256
    ecs.gn5-c8g1.2xlarge 8 60 1*NVIDIA P100 3 500
    'ecs.gn5-c8g1.4xlarge 16 120 2*NVIDIA P100 5 500
    ecs.gn5-c28g1.7xlarge 28 112 1*NVIDIA P100 5 500
  • Images
    Image name Description
    py27_cuda90_tf1.12_ubuntu Supports GPU-based TensorFlow 1.12.
    py27_cpu_tf1.12_ubuntu Supports CPU-based TensorFlow 1.12.
    py36_cuda101_tf2.1_torch1.4_ubuntu Supports GPU-based TensorFlow 2.1 and PyTorch 1.4.
    py36_cpu_tf2.1_torch1.4_ubuntu Supports CPU-based TensorFlow 2.1 and PyTorch 1.4.
    py36_cuda100_paitf1.12_alios Supports GPU-based PAI-TensorFlow 1.12.
    py36_cpu_paitf1.12_alios Supports CPU-based PAI-TensorFlow 1.12.
    py36_cuda100_tf1.15_ubuntu Supports GPU-based TensorFlow 1.15.
    py36_cpu_tf1.15_ubuntu Supports CPU-based TensorFlow 1.15.

GPU On-sale Edition

GPU On-sale Edition is upgraded from DSW 1.0. This edition is developed based on the Apsara big data platform of Alibaba Cloud and greatly reduces the costs of running DSW instances. However, this edition does not support Internet access, root permissions, or sudo operations. Purchase with caution. The following section describes the features, instance types, and supported zones of this edition.
  • Features
    • Supports real-time resource monitoring. Monitors CPU or GPU utilization during algorithm development.
    • Supports a variety of data sources, such as MaxCompute, OSS, and Apsara File Storage NAS.
    • Allows you to write and execute SQL statements.
    • Supports multiple resource types, including a variety of CPUs and GPU models.
    • Allows you to switch among different types of resources, which reduces resource usage costs.
    • Provides built-in big data development packages and algorithm libraries, and allows you to install third-party libraries.
  • Instance types and zones
    Instance type Region
    P100 China (Beijing)
    M40 China (Shanghai)

Explorer Edition

Explorer Edition is upgraded from DSW 2.0. This edition is developed based on cloud-native technologies of Alibaba Cloud, such as Docker and Kubernetes. It provides open and AI-assisted development environments for you to train models with high elasticity. Explorer Edition is free of charge. You can click Explorer Edition to try it. For more information about how to use this edition, see DSW user guide.