Persist AI Model Runner llama-server.exe for more than 5 minutes

Docker Desktop Model Runner
1

Running:
Windows 11 24H2, OS Build 26100.4652
Docker Desktop 4.43.2

The issue/feature:
The model I use is large and takes time to load in memory. If I don’t ask it a question before the end of 5 minutes, it unloads. It will happily reload after, but now I have to wait for it to reload. Is there a keepalive setting somewhere to keep the model alive in memory for an indeterminate amount of time?

Steps to replicate issue/feature:

  1. Open Docker Desktop
  2. Select [Models] tab
  3. Select an existing model to open the chat window
  4. Ask a question …
    … the com.docker.llama-server.exe starts and loads the model into memory, answers the question
  5. Let sit idle and after 5 minutes, com.docker.llama-server.exe unloads and the model is cleaned from memory

Tried with no success

  1. Disabled Docker Desktop Settings → Resources → Resource Saver
  2. Looked over docker *-options.json files for possible relevant parameters