Is there a way to limit the number of concurrent synthetic data outputs at a time?

[martymcfly1]

I'm using a local model and getting a number of failed outputs due to timeout issues. Is there a way to change the timeout value, or (better still) restrict the number of concurrent outputs being generated?

[scosman]

Not anything you can easily configure.

Most local AI hosts should have a queue. You might be able to set the max concurrency there, and have them queue up instead of fail. Example OLLAMA_NUM_PARALLEL env var for Ollama.

But if it's queuing fine, and it's just a timeout in Kiln, the only fix for now is to retry once it fails (not ideal, I know).

[scosman]

FYI: this may be in the next release