Start SR from a given expression

[dfgoe55]

Hi,

first of all I apologise if this question has been already answered. I used PySR on a large dataset and after some time I've got a nice symbolic expression. However, this is not 100% satisfactory and I want to do more SR to improve on that.

I do not have the pickle file or the hall_of_fame.csv as I run other regressions after the lucky one so I believe everything has been overwritten. I just have the symbolic expression. Is there a simple way to seed the population with that expression?

[MilesCranmer]

Have you checked your outputs folder? It shouldn't have been overwritten

[dfgoe55]

I do not have the output folder because I am running PySR in a docker, every time I turn the docker off, I think the docker kills the data. Is there an easy way to start from a known expression?

Anyway I am getting similar results now so I would have the output folder as soon as it finishes. My understanding is that the easiest is to use the procedure you described here but I didn't quite get how to do that. Could you please explain me? I just want to start from where the last SR left off.

[MilesCranmer]

The approach there requires you to have access to the saved files (if you meant the loading from a checkpoint). For docker it’s easy to mount a directory that you can save runs to.

Alternatively you should check out TemplateExpressionSpec.

A future option is MilesCranmer/SymbolicRegression.jl#469 but it’s not ready yet, unless you are ok to get your hands dirty.