Releases: lucidrains/g-mlp-gpt
Releases · lucidrains/g-mlp-gpt
0.0.15
0.0.14
prepare for local attention (to be paired with local SGU)
0.0.12
add causal tiny attention, for use with nonlocal SGU
0.0.11
allow for customizable gating activation in SGU, fix bug with padding
0.0.10
allow for customizable gating activation in SGU, fix bug with padding
0.0.9
assert
0.0.8
cache causal masks
0.0.7
easier hyperparams
0.0.6
use feedforwards for sibling block in reversible net, to save on memory
0.0.5
fix remaining bugs with axial folds