[BUG]: Warm start from a loaded model not using the previous best equations

[romanovzky]

What happened?

Hi, I was fitting a model that hanged. So I decided to load it to continue the run

model = PySRRegressor.from_file(run_directory=PATH)
model.warm_start = True
model.fit(X_train, y_train, variable_names=feature_columns)

however, I noticed that it was not continuing from where it left of, but started a new run and rewrote the CSV file with new (and worse) equations. I then tinkered a bit to try to understand, and did this

model = PySRRegressor.from_file(run_directory=PATH)
model.warm_start = True
old_run_id = model.run_id
model.run_id = model.run_id+"-2"
model.run_id_ = model.run_id_+"-2"
model.fit(X_train, y_train, variable_names=feature_columns)

and it starts a new run, ignoring the previous hall of fame, but at least it doesn't re-write the previous run.

Am I missing something?

Version

1.3.1

Operating System

Linux

Package Manager

Other (specify below)

Interface

Jupyter Notebook

Relevant log output

Extra Info

package manager: uv

[MilesCranmer]

Might be related to #816?

If you can make a MWE it would be much appreciated. I'm not able to reproduce this at the moment, though it certainly seems like there's some sort of bug here.