Saving UMAP model with more than 4096 samples results in crashing

Nobody has claimed this yet.

I am trying to save a UMAP model that got trained with more than 4096 samples and 420 features each. This results in crashing either the jupyter notebook or the editor I use. When I try to load the saved file I get an EOF Error probably because the file is corrupted since juypter notebook crashed before fully saving the file. This happens when using joblib or pickle. The versions I use are:

Python 3.10.2
umap-learn==0.5.3
numba==0.55.1
pynndescent==0.5.6

#Code
#UMAP instantiating and fitting, vectors is array of vectors
reducer = umap.UMAP(random_state=42, min_dist=0.0,n_neighbors=20).fit(vectors)
#saving via pickle
f_name = 'saving_example.sav'
pickle.dump(reducer, open(f_name, 'wb'))
#or saving via joblib
joblib.dump(reducer, f_name)

#in another file
#load model
#pickle
f_name = 'saving_example.sav'
loaded_model = pickle.load((open(f_name, 'rb')))
#joblib
loaded_reducer = joblib.load(f_name)

The crashing also happens when trying to transform new/different data to a model that was fitted using more than 4096 samples.
Can somebody help please?

Start by reproducing the failure with the shown UMAP fit using more than 4096 samples, then test the pickle.dump, joblib.dump, pickle.load, joblib.load, and transform entry points. Done means saving and loading the fitted model, and transforming new data, no longer crashes for this dataset size.

Written by the indexing model from the issue text.