Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
A transformer model replacing the attention matrices by sparse matrices to go faster.