It seems to have some problem with shuffle function in balance.py about pygta5

[Sentdex]

Yeah, I came across this, I honestly have no idea what is happening there, makes no sense to me why that'd be happening. The way I "solved" it was to put the data into a pandas dataframe, and shuffle that. Doing that gives the results one would expect.

A few models were trained by me on the old method and it still appears to work, and I looked into the data itself and it appeared correct, but the behavior of the counts still baffles me.

Here's the code I have been using to do the shuffling:

        train_data = np.load('training_data-{}.npy'.format(i))
        df = pd.DataFrame(train_data)
        df = df.iloc[np.random.permutation(len(df))]

I will soon be updating the repo as well with these changes.