Conversation
Jutho
changed the title
Codecov Report❌ Patch coverage is
... and 16 files with indirect coverage changes 🚀 New features to boost your workflow:
|
|
I keep struggling with making the QR and LQ tests pass; I thought that it was an RNG / Float32 combination that was causing the Mooncake-LQ tests to fail, and after making some changes now the Enzyme tests fail for a bunch of combinations. |
lkdvos
left a comment
lkdvos
left a comment
There was a problem hiding this comment.
I have to say that I also can't really spot something immediately, the changes all look reasonable to me
Jutho
force-pushed
the
jh/testimprovements
branch
from
df94174 to
b235cef
Compare
|
The errors seem to fail due to type instabilities in |
|
Yes, I have no idea what happened but somehow this only started failing after I merged the PR, even though the tests all passed during the PR. I'll investigate today, but we can safely ignore that here. |
Jutho
changed the title
|
Ok, then I think this is ready for another round of review. The CI is of course annoying, with type stability failing on LTS and JET code quality failing on julia 1. |
lkdvos
left a comment
lkdvos
left a comment
There was a problem hiding this comment.
Left one small comment, and we are now definitely mixing unicode subscripts and regular subscripts all over the place, but otherwise this looks great to me!
True, I'll try to fix this and remove the commented code. |
lkdvos
force-pushed
the
jh/testimprovements
branch
from
8658b5c to
740d473
Compare
2 participants
Some changes as discussed. Somehow the AD correctness tests fail for Mooncake LQ, whereas they don't do for Mooncake QR (which uses the same approach), nor for Enzyme or ChainRules (I think). I haven't figured out what is going on yet, but maybe some reviewer has an idea.