Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
|
|
|
|
The number of rigid groups overall is usually greater than the number of
rigid groups in the current collection. Therefore, the "previous rg"
function wrapped round to the wrong place.
|
|
|
|
|
|
|
|
|
|
Need a better solution here: fix them or delete them
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Scaling needs the log residual, PR needs the linear one
|
|
The gradients are correct, the test way wrong.
|
|
|