Releases: lucidrains/q-transformer
Releases · lucidrains/q-transformer
0.0.10
complete the conservative regularization loss, align some hyperparame…
0.0.9
may be a good project to try tensortype
0.0.8
discounting was not done correctly for n-step
0.0.7
0.0.7
0.0.6
per-fect attention
0.0.5
add dueling heads as an option for the q robotics model
0.0.4
layernorm is enough
0.0.3
get action functions off the q model for interacting with environment…
0.0.2
deal with batchnorm and distributed, but plan on removing batchnorm
0.0.1
vanilla q-learning for a robot transformer taking single action