Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
Attention mechanisms
Most transformer models use full attention in the sense that the attention matrix is square.