etpeterson
commented
7 months ago Added support for >1D inputs to the wrapper. Would be appropriate if it was tested by actually fitting an image.
It is quite slow though. I'm getting 300 fits/s. But that's probably because of all the ways we are calling functions through the wrapper code?
Regarding testing, what about a small multi-dimensional test? Not a whole image, but a 10x10 synthetic patch?
And regarding speed, what algorithms? In the current testing pipeline we record speeds and I see your algorithms going at about that pace. Could it help to cache the inputs in some way? It just makes me think of this decorator, but I can't think of how to use it at the moment. https://docs.python.org/dev/library/functools.html#functools.lru_cache
Added support for >1D inputs to the wrapper. Would be appropriate if it was tested by actually fitting an image.
It is quite slow though. I'm getting 300 fits/s. But that's probably because of all the ways we are calling functions through the wrapper code?