2 ms·
It's "keeping" 1 byte worth of information from each input example. The SD models are 5GB together, and the dataset 2.3B images.
by visarga 4y ago
It's "keeping" 1 byte worth of information from each input example. The SD models are 5GB together, and the dataset 2.3B images.