Can a base DLAMI be customized and used for both GPU and CPU backed instances?

It is possible to create a DLAMI that can be used for both GPU and CPU backed instances. However, it may require some additional customization and configuration.

When creating a DLAMI for GPU instances, it typically includes all the necessary GPU drivers and libraries that are required for deep learning tasks. These drivers and libraries may not be needed or compatible with CPU instances.

To create a DLAMI that can be used for both GPU and CPU instances, you will need to customize the image to include the necessary drivers and libraries for both types of instances. You may also need to modify the configuration to ensure that the DLAMI is optimized for both types of instances.

Another option is to create separate DLAMIs for GPU and CPU instances, but use a shared configuration and setup for the software and tools that are installed. This can help to minimize the amount of customization that is required and make it easier to manage and maintain the DLAMIs.

In terms of specific DLAMIs that support both CPU and GPU instances, the AWS Deep Learning AMI for Ubuntu 18.04 includes both TensorFlow and PyTorch support for both CPU and GPU instances. However, it may still require some customization to ensure optimal performance and compatibility with your specific use case.