Is it correct that even if I have for example, 10k audio samples to train from, the model will only use a slection of 4096 to generate training data, as shuffle buffer size is set to that value, and in a training session it does not appear to load any more samples once the 4096 have been loaded?
I am using a set of 49k samples, and have changed shuffle buffer size to match. Do I actually need to do this?
Is it correct that even if I have for example, 10k audio samples to train from, the model will only use a slection of 4096 to generate training data, as shuffle buffer size is set to that value, and in a training session it does not appear to load any more samples once the 4096 have been loaded?
I am using a set of 49k samples, and have changed shuffle buffer size to match. Do I actually need to do this?