3 ms·
Just the sheer size of JFT (latest versions I heard approach 1B images), so it's probably impractical to train on it till overfitting.
by quantumofalpha 6y ago
Just the sheer size of JFT (latest versions I heard approach 1B images), so it's probably impractical to train on it till overfitting.