2 ms·
The parameters of neural networks are involved in matrix multiplies, so it's an O(n^2) operation per layer. 32 tokens/ second sounds great
by mikewarot 3y ago
The parameters of neural networks are involved in matrix multiplies, so it's an O(n^2) operation per layer.
32 tokens/ second sounds great