Seeking Methods to Track Loss, Complexity, and Score in PySR Iterations Without Custom SymbolicRegression.jl

[edwardaud]

I am using PySR for symbolic regression and need a simpler way to log loss, complexity, and score per iteration without customizing SymbolicRegression.jl. Does anyone have a generalized approach or suggestions for this?

[MilesCranmer]

You can use this strategy: #439 (reply in thread). But I would actually use more populations, and maybe 2 iterations per run so that migration happens.

And then when I finally finish adding TensorBoard, that would be even easier: MilesCranmer/SymbolicRegression.jl#277

Also see progress on a UI in #589