[Feature] Job Concurrency overrides for Remote Machine Learning servers

[krunkster]

I have searched the existing feature requests, both open and closed, to make sure this is not a duplicate request.

• Yes

The feature

As an Immich server admin using a low-power server I want to limit job concurrency on the server to minimal levels to prevent over taxing the server. However, when my remote machine learning server is available I would like to automatically increase job concurrency that utilizes that server to make full use of the remote server's capabilities.

This to me implies per remote server configuration overrides, but I'm sure there are other ways to slice this problem.
Regardless, I acknowledge this is likely a bit of a configuration management / nesting nightmare, but hopefully the request makes sense.

Platform

• Server
• Web
• Mobile

[mertalev]

Hi, you can set MACHINE_LEARNING_REQUEST_THREADS on the low-power server's ML container so it can only process that many assets at a time even if the server's job concurrency is higher. This lets the remote ML container still see a higher job concurrency so it can be utilized effectively.