Set up multi-user mode (admin) - AWS DeepRacer

Set up multi-user mode (admin)

With a multi-user account setup, organizers (such as account administrators) can provide participants access to AWS DeepRacer service under their account ID. They can also set usage quotas on participants' training hours, monitor spending on training and storage, enable and disable training, and view and manage models for every user in their account from the AWS DeepRacer console.

Multi-user mode is particularly useful for large events with multiple participants who don't have individual AWS accounts. Instead of creating and managing accounts for each participant in an event, an AWS DeepRacer administrator can host all of their sponsored participants through a single AWS account.

In multi-user mode, sponsored participants can compete and train without incurring any of their own costs. Their training and storage charges are billed to the sponsoring multi-user AWS account billing. If an administrator stops sponsoring participants' usage, participants keep their racer aliases and profiles.

Multi-user stakeholders

This walkthrough refers to the following typical multi-user stakeholders for setting up and using multi-user mode.

  • AWS administrator for IAM/SSO configuration. The AWS administrator for IAM/SSO configuration sets up IAM or SSO for the AWS DeepRacer administrator and for participants to use multi-user mode. The AWS administrator for IAM/SSO has IAM and SSO administrator permissions. For information about creating IAM users, see Create IAM users.

  • AWS DeepRacer administrator. The AWS DeepRacer administrator manages AWS DeepRacer participants' sponsorship and can pause and resume sponsorship, delete models and artifacts, configure and host virtual races, and enable and disable multi-user mode. The AWS DeepRacer administrator has AWSDeepRacerAccountAdminAccess permissions.

  • AWS DeepRacer participant. AWS DeepRacer participants are invited to participate in events under an administrator's AWS account in multi-user mode. Participants have AWSDeepRacerDefaultMultiUserAccess permissions to train, evaluate, and store models in the sponsor's account. Participants also configure their racer profile, enter virtual races, and download their models for deploying on a physical AWS DeepRacer vehicle.

In this walkthrough, you perform the following steps:

  • Step 1. Perform prerequisites.

  • Step 2. Activate multi-user mode on your AWS DeepRacer account.

  • Step 3. Invite participants.

  • Step 4. Set usage quotas.

  • Step 5. Monitor usage of your sponsored participants.

Step 1. Prerequisites for AWS DeepRacer multi-user mode

Complete the following prerequisites for multi-user mode:

Set up your account with AWS DeepRacer admin permissions for multi-user

To set up as an AWS DeepRacer admin for multi-user mode, you need to have the IAM AWS DeepRacer administrator policy, AWSDeepRacerAccountAdminAccess, attached to your user, group, or role. Depending on your organization, you may set yourself up with the administrator policy by using the console to create a user or role and attaching the required IAM policy, or you may have your IT administrator provide it. For information about the required administrator policy, see AWSDeepRacerAccountAdminAccess. For more information about IAM policies, see Access Management in the IAM User Guide.

Provide AWS console access to your sponsored participants

To provide racers you sponsor with access to the AWS DeepRacer console, we recommend using standard AWS authorization protocols such as AWS IAM Identity Center (successor to AWS Single Sign-On) or AWS Identity and Access Management. You can also provide access through your organization’s preexisting SSO. When participants log in to the AWS DeepRacer console using the credentials you provide, they are prompted to create an AWS player account to log in and access the AWS DeepRacer console under your AWS account. For more information about AWS Player accounts, see AWS Player accounts.

  1. Create an IAM username and password for each participant. See Create IAM users.

  2. Grant each participant the permissions in AWSDeepRacerDefaultMultiUserAccess. For more information, see AWS managed policies for AWS DeepRacer.

  3. Email participants with IAM usernames and passwords as well as a link to the console. Using the provided link and entering their IAM usernames and passwords, participants can access the console. For information about creating IAM users, see Create IAM users.

  1. Open the IAM Identity Center console at https://console.aws.amazon.com/singlesignon/, create a custom permission set, and assign users to the account. For more information, see Permission sets.

  2. When creating the custom permission set, provide the following values:

  • Relay state: https://console.aws.amazon.com/deepracer/home?region=us-east-1#getStarted

    Note

    The Relay state redirects participants within the account to a specified URL; in this case, it directs them to the AWS DeepRacer console.

  • AWS Managed policies: AWSDeepRacerDefaultMultiUserAccess

After you have fulfilled the prerequisites, you are ready to activate multi-user mode and invite participants to race through your account.

Step 2: Activate multi-user account mode

After you have set up your AWS DeepRacer admin account and granted console access and permissions to your sponsored participants, you can activate multi-user mode on your AWS DeepRacer account.

Note

By default, there are limits on accounts sponsoring participants in multi-user mode. For more information, see the section on account limits in Monitor Usage.

  1. In the left navigation pane, navigate to Multi-user management and the Setup page.

  2. In Enable multi-user account mode, turn on Enable multi-user mode.

  3. In the Enable multi-user mode dialog box, select the checkboxes to confirm that your sponsored participants have required access and permissions.

  4. Choose Enable multi-user mode.

    When you meet the prerequisites and activate multi-user mode, each of your sponsored participants can train models with all training and storage charges billed to the administrator's AWS account. By default, a participant has a limit of 2 concurrent models.

Disable multi-user account mode

Disabling multi-user mode ensures that no new profiles can be created under your administrator account and the profiles of previously sponsored participants are no longer visible on the administrator's account. Participants are no longer prompted to log in to their AWS player accounts and can't access or train models created under the administrator's account.

The administrator can download, save, and import sponsored participants' models.

  1. Navigate to Multi-user management and the Setup page.

  2. In Disable multi-user account mode, choose Disable multi-user mode.

  3. In the Disabling multi-user mode dialog box, select the checkbox to confirm that you want to disable multi-user mode. Choose Disable multi-user mode.

    Multi-user mode is disabled.

    Note

    All models created under a sponsoring AWS multi-user account persist and model storage costs continue on the AWS account until models are deleted.

Step 3: Invite participants to be sponsored

You can invite participants to train and race as sponsored participants by using the provided email template.

To invite participants

  1. In the left navigation pane, navigate to Multi-user management and the Setup page. Under Set up multi-user mode in the Invite users section, choose View invite template.

  2. Copy the email template that appears into your email client application and use it to craft an email to send to the participants you want to invite to be sponsored. If you are using your company's existing SSO, you can include a SSO URL for your participants to use. Alternatively, you can provide IAM credentials for the participants to use to log in to the AWS console.

Step 4: Set usage quotas

After your sponsored participants have received their invitation email and have created their profiles under your account, they appear in the Sponsored users list in the Monitor Usage screen. In this screen, you can then set usage quotas on the number of available training hours and models for sponsored participants. By setting quotas, you can control costs per participant under your account and ensure that participants can't exceed their usage quota. You can also increase or decrease usage quotas as needed to provide sponsored participants with the hours they need to effectively train an AWS DeepRacer model.

Note

By default, sponsored participants in multi-user mode receive 5 hours of training time.

To edit usage quotas for sponsored racers

  1. In the left navigation pane, navigate to Multi-user management and the Monitor usage screen. In the Monitor usage screen in Sponsored users, select the participants for whom you want to set limits. Choose Actions to open the dropdown list and choose Set usage quotas.

  2. In the Set usage quotas popup, enter the Maximum training hours and Maximum model count for the users you selected. Choose Confirm to keep your changes or Cancel to discard them.

Step 5: Monitor usage

You can monitor the usage of your sponsored participants, including estimated spending and training model hours. You can also pause sponsorship of participants, delete models, and view summaries of usage. You perform all tasks related to monitoring usage in AWS DeepRacer Multi-user management in the Monitor usage page.

All information about expenses for sponsored racers is an estimate only and should not be used for budgeting or cost accounting purposes. Estimates are in USD and do not reflect any special pricing. For more information about pricing, see Pricing.

Account limits for multi-user mode

By default, a sponsoring account in multi-user mode has the following limits which are shared among all sponsored profiles :

  • 100 concurrent training jobs

  • 100 concurrent evaluation jobs

  • 1000 cars

  • 50 private leaderboards

To adjust these limits, contact Customer Service.

To view an estimate of spending

On the Monitor usage page, under Monitor usage, you can view an estimated summary of your participants' usage.

To set up billing alerts

You can set up billing alerts for your account. Billing alerts help you to keep up to date on spending. For more information, see Billing.

To pause sponsorship

You can pause sponsorship of a single participant, multiple participants, or all participants. When you pause sponsorship, your sponsored participants can't create new models or train models under your account. Training that is in progress runs through to completion and is included in estimates for spending. You can resume sponsorship at any time. Participants whose multi-user access has been paused can still view their models and post models to leaderboards, but they can't perform any cost-generating activities.

  1. On the Monitor usage page, under Monitor usage, in the Sponsored users section, select the users for whom you want to pause sponsorship.

  2. Choose Pause sponsorship.

  3. In the Pause sponsorship dialog box, choose Pause sponsorship to pause sponsorship. Choose Cancel if you decide that you don't want to pause sponsorship.

To resume sponsorship

You can resume sponsorship of racers for whom you have paused sponsorship.

  1. On the Monitor usage page, under Monitor usage, in the Sponsored users section, select the racers for whom you want to resume sponsorship.

  2. Choose Resume sponsorship.

To view racers' models

  • On the Your Models page, under Models, you can view your models and your users' models.

Next steps

After you have set up and activated multi-user mode, you can take the following steps:

  • Create a community race.

  • Request an AWS DeepRacer workshop.

Create a community race

Community races provide an exciting way for your sponsored participants to experience reinforcement learning.

You can create community races and invite your sponsored participants.

For more information, see Create a Virtual Community Race: A Quick Start Guide .

Request a workshop

You can request a workshop to learn more about AWS DeepRacer with a 60-minute online or in-person workshop.

For more information, see Workshop.