Restore old MLL fit #320
Restore old MLL fit #320
Conversation
|
Hi @Scienfitz, thanks for the change, looking good 👍🏼 I suggest that we wait with the merge until the scaling is in, to have this as a last sanity check before the new release (please keep the files so that we can rerun again once everything is in place!) |
|
@AdrianSosic please can I merge this? I need to move forward with my items, and the block here seems artificial as the PR has nothing to do with running the tests which you or I are perfectly capable of doing independently of this PR |
Scienfitz
force-pushed
the
fix/mll_fit
branch
2 times, most recently
from
15f17e4
to
a68c2a7
Compare
There was a problem hiding this comment.
Ok, if it helps you to progress faster, let me accept. But I think the tests are related to this PR because who nows if stuff still works as shown here once all changes are in? So let's definitely have last final test before the release, which potentially requires adjusting the approach once again
Co-authored-by: AdrianSosic <[email protected]>
use torch fitting with 100 steps again until we finetune this better
here some pictures with the new fit indicating previous results are there again
tobis notebook with this PR
vs old (non torch) fitting
