All Products
Search
Document Center

Platform For AI:DSW overview

Last Updated:Jan 20, 2025

Data Science Workshop (DSW) of Platform for AI (PAI) is a one-stop Integrated Development Environment (IDE) for AI tailored for algorithm developers. DSW integrates multiple development environments, such as Notebook, VSCode, and Terminal, for coding, debugging, and task running. DSW provides various heterogeneous computing resources and open-source images and supports mounting of datasets of the Object Storage Service (OSS), File Storage NAS (NAS), and Cloud Parallel File Storage (CPFS) types. You can manage the lifecycle of DSW instances and use DSW for development in an easy and efficient manner.

Advantages

  • Flexibility and ease of use

    • DSW provides built-in development environments, such as Notebook, VSCode, and Terminal, to meet various development requirements.

    • DSW provides images of multiple open-source frameworks such as PyTorch and TensorFlow, and supports custom images.

    • DSW provides various heterogeneous computing resources, including public resource groups, dedicated resource groups, and Lingjun resources. You can flexibly configure and manage resources in DSW.

    • DSW supports the writing and execution of R language and SQL statements on top of Python.

  • One-stop service

    • DSW allows you to mount file systems, such as OSS, NAS, and CPFS file systems, and access MaxCompute data.

    • DSW provides Deep Learning Containers (DLC) and Elastic Algorithm Service (EAS) tools to implement a full AI development pipeline from data processing, coding, debugging, model training, to model deployment.

    • DSW provides the AI coding assistant Tongyi Lingma to improve coding efficiency.

  • Fine-grained management

    • DSW allows you to configure scheduled stop for an instance or auto stop for idle instances to reduce costs.

    • DSW provides real-time monitoring of CPU, GPU, and memory usage to help you analyze the resource usage in real time.

    • A workspace administrator can allocate global resources and configure resource reclamation strategies.

  • Scenario-based tutorials

    • DSW provides Notebook Gallery as a content platform for developers. You can use the tutorials for large language model (LLM) and AI content generation-related industries in Notebook Gallery to quickly get started with development.

Usage process

Instance creation and access

Grant the permissions that are required to use DSW

Before you use DSW, grant the operation account the required creation and development permissions.

Create a DSW instance

During instance creation, you can select a resource type, attach a dataset, and select a custom image based on your business requirements.

Access a DSW instance

You can access a DSW instance in the PAI console in a simple manner and use the features of DSW.

Connect to a DSW instance

DSW enables you to connect to an instance remotely by using SSH through an on-premises terminal or VSCode. This facilitates the running and debugging of your on-premises code in the cloud.

Instance configuration and management

Manage DSW instances

You can manage and change the lifecycle and configurations of an instance, such as configuring shutdown policies and optimizing instance costs.

Mount datasets or OSS paths

To expand the storage space of an instance, persist storage data, or read data files, attach a dataset to the instance and mount an OSS directory.

DSW network configuration

To use an instance in a VPC, improve the data upload or download speed, or manage public access, you need to configure network parameters for the instance.

Configure RAM roles for a DSW instance

You can associate a RAM role with an instance and access other cloud services from the instance by using a Security Token Service (STS) temporary credential without the need to configure AccessKey pairs. This reduces the risk of key leakage.

Model development and deployment

Using Tongyi Lingma for development

DSW features the built-in AI coding assistant Tongyi Lingma to provide various capabilities, such as code completion and optimization and intelligent Q&A. This facilitates efficient development.

Read data from and write data to OSS and MaxCompute

You can read OSS or MaxCompute data files from an instance by using an API or SDK.

File upload and download

You can transmit data and models between on-premises machines and instances.

Custom services access configuration

After you build a model, you can use this feature to enable DSW to provide services over the Internet.

Deploy a model

If you want to call the model that you recently built from other applications or perform elastic scaling, version control, or resource monitoring, you can deploy the model as an online service.

Advanced features of DSW

Notebook Gallery

Notebook Gallery provides various notebook cases, including cutting-edge areas such as LLM and AI-generated content (AIGC), and popular models, which you can run with a few clicks and optimize.

TensorBoard: training visualization

The TensorBoard plugin is provided to display the metrics and relevant information during model training in a visualized manner.

Install the R kernel in DSW

DSW is integrated with open-source JupyterLab. You can install the R kernel on a DSW instance to run R scripts for data analysis.

Billing

Compute instance

You can select public resources and dedicated resources such as general computing or Lingjun resources. Different resources use different billing methods.

Resource type

Billing method

Billable item

Billing rule

How to stop billing

Public resources

Pay-as-you-go

Service duration of a DSW instance (the duration for which a DSW instance occupies public resources)

If you use public resources to create a DSW instance, you are charged based on the service duration of the DSW instance.

  • Stop a DSW instance.

  • Delete a DSW instance.

Important

You must manually stop an instance or configure scheduled stop. For more information, see Manage DSW instances.

Dedicated resources (general computing or Lingjun resources)

Subscription

Number of nodes of an instance type and purchase duration

If you purchase subscription dedicated resources, you are charged based on the number of nodes of an instance type and purchase duration. For more information, see Billing of AI computing resources.

None.

System disk

Billing method

Billable item

Billing rule

How to stop billing

Pay-as-you-go

System disk capacity and usage duration

After you expand the system disk capacity, you are charged for the capacity that exceeds the free quota and usage duration.

Delete a DSW instance.

For more information, see Billing of DSW.