When tfserving switches model versions, a 2-3 minute service unavailability occurs

[momodagithub]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):centos
• TensorFlow version and how it was installed (source or binary):2.8.3
• TensorFlow-Recommenders-Addons version and how it was installed (source or binary):0.6.0
• Python version:3.9
• Is GPU used? (yes/no):no

Describe the bug

This behavior isn't exactly a bug. When I use a custom tfserving image compiled with tfra to provide model services, I find that the tfserving service becomes unavailable for 2-3 minutes when switching model versions. How can I resolve this issue?

it seems that setting num_unload_threads=1 does not work