Data Science Workshop (DSW) of Machine Learning Platform for AI provides heterogeneous computing resources. The computing resources include GPUs of different models such as V100, P100, and M40, and CPUs of different specifications. You can orchestrate computing resources of pay-as-you-go DSW instances for different types of jobs. You can use CPU resources for coding, which is cost-effective. When you need to train models, you can switch to GPU resources.

Resource specifications

Resource type Resource specification Supported region
pai.medium.1xv100 NVIDIA V100 China (Beijing), China (Shanghai), China (Hangzhou), and China (Shenzhen).
pai.medium.1xp100 NVIDIA P100 China (Beijing), China (Shanghai), China (Hangzhou), China (Shenzhen), and Singapore (Singapore).
Note A DSW instance deployed in the China (Beijing) region supports up to eight GPUs.
pai.medium.1xm40 NVIDIA M40 China (Shanghai).
pai.large.2core4g 2 CPU Core + 4 GB China (Beijing) and China (Shanghai).
pai.xlarge.4core8g 4 CPU Core + 8 GB China (Beijing) and China (Shanghai).
pai.2xlarge.8core16g 8 CPU Core + 16 GB China (Beijing) and China (Shanghai).
pai.4xlarge.16core32g 16 CPU Core + 32 GB China (Beijing) and China (Shanghai).
pai.6xlarge.24core48g 24 CPU Core +48 GB China (Beijing) and China (Shanghai).

Pay-as-you-go instances are billed based on the duration that the instances have been running. The price varies based on the type of resource that the instance uses. After you switch to another type of computing resource, the system starts to charge you based on the new computing resource other than the previous one. For more information about how you are charged for using different types of resources, see PAI-DSW计费说明.

Switch computing resources

If no sufficient resources are available for model training, you can switch to another type of computing resource. You can switch the computing resources of a pay-as-you-go DSW instance between different specifications of CPUs, between CPUs and GPUs, and between different models of GPUs.

Click Switch to switch to another type of computing resource. After you select the resource type to which you want to switch, Switching is displayed on the top of the panel. The switchover is complete when the selected resource is displayed on the top of the panel. It takes 1 to 2 minutes to complete the resource switchover. You must re-run the code after the switchover is complete.
Note
  • Do not perform other operations during the switchover.
  • After the switchover is complete, the system starts to charge you based on the new resource. Please pay attention to your bill statements. For more information about how you are charged for using different types of resources, see PAI-DSW计费说明.
  • If the page is not updated or shows an error message after you perform the switchover, manually refresh the page.
  • We recommend that you do not frequently switch computing resources unless necessary.