0.11.3

Merge pull request #41 from lucidrains/pw/fix-normalization

fix normalization for fast cuda version of causal

0.11.2

fix bug with non-cuda linear causal attn

0.11.1

Update setup.py

0.11.0

set projection matrix on init

0.10.0

bump for @norabelrose PR

0.9.2

change initial value of rezero parameter to 1e-3 per @fcampagne sugge…

…stion

0.9.1

fix device error when redraw projection is false

0.9.0

make performer work as encoder / decoder with cross attention

0.8.1

allow for no local attention heads

0.8.0

turn on Shaws relative positional encoding for local attention