I mean to reduce the API calls without reducing the scalars that are logged, e.g. by sending less frequent batched updates.

Understood,

In my current trials I am using up the API calls very quickly though.

Why would that happen?
The logging is already batched (meaning 1API for a bunch of stuff)
Could it be lots of console lines?

BTW you can set the flush period to 30 sec, which would automatically collectt and batch API calls
https://github.com/allegroai/clearml/blob/25df5efe74972624671df2ae97a3c629eb0c5322/docs/clearml.conf#L196