Releases: lucidrains/se3-transformer-pytorch
Releases · lucidrains/se3-transformer-pytorch
0.5.7
add absolute positional embedding to se3 transformers
0.5.6
apply rotation to values as well, for rotary embeddings
0.5.5
apply rotation to values as well, for rotary embeddings
0.5.4
fix bug where rel dist may contain masking values used for ranking ne…
0.5.3
add rotary positional embedding, for either sequential position or re…
0.5.2
turn on final normalization in the presence of reversibility
0.5.1
fix global features under the setting of one headed key values
0.5.0
add ability to tie the key / values, for 2x memory savings
0.4.0
add feature where one can share one key / value head across all heads…
0.3.15
add ability to use linearly proj keys in AttentionSE3