4 ms·
This model was trained on 6T tokens and has 256k embeddings, quite different than a gpt2 model comparable in size.
by GaggiX 1y ago
This model was trained on 6T tokens and has 256k embeddings, quite different than a gpt2 model comparable in size.