crypto data features in x and future 1/2/6/12/24h ret in y as target.
x is numpy array in shape (33, 769), 33 top liquid coins, 769 features:
h, w = 33, 769
numpy.memmap(x_path, shape=(h, w), dtype=np.float32, mode='r')
y is numpy array in shape (33, 5)
