Takes ages because it is pickling and unpickling these numbers.
There are a few things slowing this down:
Eval uploads should be batched or async streamed.
Even if the eval is stored locally, it needs to be unpickled from the local cache. There should be special support for basic scalar result types like strings and numbers.
There should be an @cache in-mem cache too. At least a weakref system.
Takes ages because it is pickling and unpickling these numbers. There are a few things slowing this down:
@cache
in-mem cache too. At least a weakref system.