TensorFlow Serving warmup

TensorFlow Serving allows warming up models using a file in assets.extra directory of the model directory. I was looking for a way to put tfserving K8S Pod in ready state (using readiness probe) only after the warmup has finished. Is there any API that shows when the model is warmed up (all records in warmup file processed)?

Hi @Mehran_S ,

To determine when a TensorFlow Serving model has finished warming up in a Kubernetes environment, you can:

  • Set a readiness probe with a delay.
  • Implement a custom health check.
  • Use external monitoring.

Hope this works ,

Thank You.