imyaman

Hello Google Cloud AutoML Vision & AutoML Vision Edge Customer,

We are writing to remind you that Cloud AutoML Vision and AutoML Vision Edge will be Generally Available (GA) later this year. Effective the GA release date, we are introducing a new version (v1) of the Cloud AutoML API and new pricing structure for online predictions with Image Classification models.

If you are an ML Kit customer that uses AutoML Vision Edge exclusively via ML Kit, please note that the ML Kit console and client libraries will remain in beta. If you have not used the Cloud AutoML Vision APIs directly via the Google Cloud Console, the following changes won't impact your usage or pricing and you may stop reading now.

Your Image Classification datasets are being migrated from the legacy Beta UI to the GA (Integrated) UI, which unifies the experience with Object Detection. Once the migration is complete, you will be redirected from the legacy UI to the Integrated UI. If you were using the integrated Human labeling service from the legacy UI, you can start using the AI Platform Data Labeling Service.

As part of the GA release:

• The v1beta1 version of the Cloud AutoML API will be deprecated in favor of the new v1 version, and you will need to migrate your applications from the v1beta1 version to the v1 version after GA.
• The v1beta1 endpoint is scheduled for deletion within six months after GA release.
• Due to the fundamental change in Image Classification pricing for GA, discounts applied to Beta pricing will not be applied to GA. You may reach out to your sales representative for more information.
• The maximum supported lifespan for custom models created in Cloud AutoML Vision and AutoML Vision Edge will change from two years to 18 months at GA.

The new pricing is summarized in the table below:

• Pricing for Cloud models will be charged using node hour as the unit.
• Corresponding free trials will receive a one-time free credit of 40 node hours per billing account lifetime rather than every month.
• The node hours represent an hour of processing on different GPU or TPU node types.
• To train Cloud models, we use 8 nodes in parallel.
• For deployment, select the number of nodes on which the model is to be deployed. You can adjust the number of nodes when you deploy your model.

When your projects are migrated to the GA (Integrated) UI, your Classification models are shown as deployed with ‘0 nodes’. Any model deployed with ‘0 nodes’ will continue to serve online predictions from the old service, but will be excluded from our Service Level Agreement offered at GA. If no action is taken, these deployments will incur charges of $1.25 per hour at GA.

Please review your Classification models in the Integrated UI and ensure none are left in their initial state of deployed with ‘0 nodes’:

1. Undeploy any models that are not required, to avoid being charged at GA.
2. Manage Classification models by redeploying or updating model’s node number to meet the traffic each model needs to serve.

Please refer to our release notes for detailed instructions.

We have identified that your projects listed below have active Cloud models which need to be reviewed. Each billing account (listed after the project ID in parentheses) will receive the one-time free credit of 40 node hours to be shared between its respective projects:

• sys5 (sys5-f9f83)

If you have any questions or require assistance managing your models, please join the cloud-vision-discuss Google group.

If you have any billing questions, please contact Google Cloud Billing Support.

Thank you for being a valued AutoML Vision customer.