0.0.14

add talking heads feature, turned on with `attn_talking_heads = True`

0.0.12

complete encoder / decoder

0.0.11

add T5 relative positional bias

0.0.10

simplify returning of embeddings or logits

0.0.9

rename layer blocks to attn layers

0.0.7

multiply by token embedding weight for logits by default

0.0.6

add feedforward GLU variant and scalenorm

0.0.5

expose dim head, show how simple gpt3 is

0.0.4

always fix head dimension at 64, so users do not trip up with imprope…

0.0.3

deploy