There has been some discussion of developing more of a consistent setup / approach for profiling the module / models, and using this as we go forward.
Recent optimizations (#299) spurred this discussion including the following comment (to keep in mind) about further work on benchmarking: #299 (comment)
For some basic record keeping, in terms of model fitting, we currently spend basically all of our time in curve_fit, with there not being much obvious opportunity to speed things up beyond what was already done in #299. See this comment (#299 (comment)) for some comments on potentially exploring different fitting approaches.
In terms of some basic / benchmarking sims / tests, there are some starting points:
I think the general idea would be to develop anything like this with or after the 2.0 release - so for now we can use this issue for any further discussion, and keep this in mind as we move towards 2.0 and beyond.
There has been some discussion of developing more of a consistent setup / approach for profiling the module / models, and using this as we go forward.
Recent optimizations (#299) spurred this discussion including the following comment (to keep in mind) about further work on benchmarking: #299 (comment)
For some basic record keeping, in terms of model fitting, we currently spend basically all of our time in
curve_fit, with there not being much obvious opportunity to speed things up beyond what was already done in #299. See this comment (#299 (comment)) for some comments on potentially exploring different fitting approaches.In terms of some basic / benchmarking sims / tests, there are some starting points:
I think the general idea would be to develop anything like this with or after the 2.0 release - so for now we can use this issue for any further discussion, and keep this in mind as we move towards 2.0 and beyond.