Segmentation fault whenever I try to train or predict

I’m running xgboost version 0.90 on macOS Big Sur with python 3.9.2. I need to use xgboost 0.90 to be compatible with what my colleagues are running. However, whenever I try to create a new model and train it, or load an existing model (trained on a different machine but with version 0.90), I get this:

Segmentation fault: 11

Does anybody know what might be happening/what I should do to fix it?

One potentially relevant point: I had 1.3.3 installed, and to downgrade to 0.90 I had to set gcc-7 in the config.mk file:

export CC = gcc-7
export CXX = g++-7

But once I did that the installation seemed to work – I get this when I import xgboost in the python interpreter:

>>> import xgboost; xgboost.__version__
'0.90'

Also, even if I set gcc-7 to be my default, I still get the seg fault when I try to train or predict. So perhaps that has nothing to do with my present issue.