Releases: lucidrains/rela-transformer
Releases · lucidrains/rela-transformer
0.0.7
0.0.6
option for relu squared activation
0.0.5
support masking for non-causal case
0.0.4
add the gated rmsnorm proposed in paper
0.0.3
causal flag
0.0.2
0.0.2a add memory key / values, and way to turn off feedforward
0.0.2
enwik8 instruct
0.0.1
working rela attention