All Products
Document Center


Last Updated: Apr 07, 2020


PAI-DSW(Data Science Workshop) is a deep learning development environment provided on the cloud for algorithm developers at different levels. DSW has integrated JupyterLab, and provided some customized features. Developers can open notebook to write code and debug directly. Also, you can read data from MaxCompute table, NAS and OSS. After tranning, models can be deployed to PAI EAS model serving by using EASCMD.

DSW Environment

The user interface contains the files section on the left side, the code editing section in the middle, and the resource search section on the right side. More details about DSW Environment, please refer to DSW Environment

Core Features

  • Real-time resource monitoring, visual display of CPU / GPU usage while algorithm development
  • Built-in common data science and algorithm libraries, and support custom installation of third-party libraries
  • Multi-source data access, including MaxCompute, OSS and NAS
  • SQL is supported in ipynb
  • Provide a variety options of resource type, including CPU and GPU.
  • Flexible switching of various resources to effectively reduce the cost of use

Supported Regions

China (Beijing) China (Hangzhou) China (Shanghai) China (Shenzhen) Singapore and India.

Billing Method

Support pre-paid (annual and monthly) and post-paid (pay-as-you-go), you can choose the payment method when creating a DSW instance. Product pricing and billing methods can refer to the document:PAI-DSW Pricing

Tips:The prepaid billing DSW instance in Beijing and Shanghai and the postpaid M40 in Shanghai do not support networking, and postpaid for other available regions support networking.

About Switching:Currently, the DSW resource switching function only supports instances of the post-paid instance。

Supported Resource Type

Resource Type Resource Details Supported Regions
pai.medium.1xv100 GPU V100 China (Beijing),China (Shanghai),Singapore
pai.medium.1xp100 GPU P100 All Regions
pai.medium.1xm40 GPU M40 China (Shanghai)
pai.large.2core4g CPU 2Core4GB All Regions
pai.xlarge.4core8g CPU 4Core8GV All Regions
pai.2xlarge.8core16g CPU 8Core16GB All Regions
pai.4xlarge.16core32g CPU 16Core32GB All Regions
pai.6xlarge.24core48g CPU 24Core48GB All Regions

Supported Frameworks

Billing Type Framework Version
Prepay TensorFlow 2.0/1.12.3
Prepay PyTorch 1.3.1
Postpay TensorFlow 2.0/1.14.0
Postpay PyTorch 1.1.0

Mounting NAS

How to add storage capacity by mounting NAS, please refer to Document.