TF: OOM when allocating tensor, but only 32 batch_size

Question

I\'m trying to vectorize all my images using EfficientNetB0 and save it for further training. This is script I\'m running:

from data import DataGenerator, vec