INFO: Created TensorFlow Lite XNNPACK delegate for CPU. Delegate this process to TPU

Server Version#: latest
Player Version#: latest

I found this row in my log: INFO: Created TensorFlow Lite XNNPACK delegate for CPU. My server has a Coral TPU that I want to use for the SONIC process. On Linux, we can configure /dev/apex_0 and /dev/apex_1. In the future, will there be an option before the process starts for the script to check if the machine has this accelerator? If so, it could train TensorFlex to use the hardware and skip using CPU resources. Thanks!

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.