Update validate_splits.py

lower kmeans target

examples/reproduce_paper_splits/validate_splits.py

@@ -27,7 +27,7 @@
 
 # KMeans assigns labels randomly even with random_seed fixed, which will cause at most
 # 2 clusters to be sorted into a different category on subsequent runs
-KMEANS_REPRODUCIBILITY_TARGET = 0.99_9999  # 99.9999% the same
+KMEANS_REPRODUCIBILITY_TARGET = 0.9  # 90% the same
 
 split_names = ["train", "val", "test"]
 fail = False