kernel gets stuck if I train/test split by 55% and 45%

Question

I am trying to train a neural net on a dataset. Everything works. There is no issue with the code if I specifiy 70% or 50% percent of the data as training and the rest as te