Create an interactive container
You can create an interactive container with the default setting using a single command, or choose a container type according to your needs.
For the detailed information of container types (including the number of GPUs and other resource sizes), please refer to TWSC Pricing.
Before we get started
- If you haven't sign up for a user account, please sign up for TWSC first.
- If you would like to operate with TWSC portal, please first sign in to TWSC portal, and go to Interactive Container Management page.
- If you would like to operate with TWCC CLI, please install TWCC CLI first, and then sign in.
Create an interactive container
- TWSC Portal
- TWCC CLI
- Go to Interactive Container Management page, and click ＋CREATE.
- Choose Image Type: You can browse and select the deep learning framework optimized by NGC, or directly search for the name of the framework in the search bar (e.g. PyTorch, TensorFlow...).
tip
- For the NGC container image type, version and packages provided by TWSC, please refer to Container Images.
- After the first selection, system would record and display the image in Recent Used block to speed up your next selection. (Clearing the browser history will also clear this record.)
- Fill in the basic information and configuration of the container, then click NEXT : STORAGE>.
info
If "Deletion Protection" is enabled, before deleting a resource, you need to disable the protection mechanism before continuing the deletion process, which can prevent the resource from being deleted by mistake and causing service interruption.
- The system will automatically attach the two directories (/home and /work) of the Hyper File System (HFS) to the container after creation for better performance when running AI computing. In addition, TWSC Cloud Object Storage is another storage option.
- Then click NEXT : ENVIRONMENT VARIABLES> to set the environment variables that meet your needs. For detailed explanation and setting examples, please refer to Setting Environment Variables.
- Then click NEXT : INITIAL SCRIPTS>. For detailed explanation and examples of how to use the initial scripts feature, please refer to this document.
- Click NEXT : REVIEW&CREATE> to check the configuration and price of the container. Then, click CREATE to confirm the creation.
Command
twccli mk ccs [-n] # Name of the container
[-itype] # Image type
[-img] # Image version
[-gpu] # Number of GPUs(Pcs)
[-envk] # Environment variables key
[-envv] # Environment variables value
info
If you do not input any parameters, the container will be created with the default settings:
| Image type, Image | Name of container | Container type |
|---|---|---|
| TensorFlow (latest environment) | twcc-cli | c.super (1 GPU + 04 cores + 90 GB memory) |
For detailed explanation and setting examples, please refer to Setting Environment Variables.
Example
- Create a container with the name
cusccs, image typeCaffe2, image versioncaffe2-18.08-py3-v1:latest, and GPU quantity2.
twccli mk ccs -itype "Caffe2" -img "caffe2-18.08-py3-v1:latest" -gpu 2 -n cusccs
caution
- Container name requirements: It must be consist of lowercase letters or numbers, the first character must be a lowercase letter, and be 6-16 characters long.
- If the setting does not meet the above requirements, the following error message will appear:
info
- For more information about the images, please refer to TWSC Container Image.
- For more information about the container types, resource ratio and pricing, please refer to TWSC Pricing.
- If available memory capacity of containers is calculated in GiB (1 GiB = 230 bytes), take c.super as an example, its available memory capacity will be 84 GiB. For more information, please refer to Memory capacity conversion for more details.
caution
- To keep the data intact, please store the important data in the Hyper File System (HFS) rather than container disks.
- After the container is created, it will continue to be billed until you delete it.