Conversation
|
(CI will fail until the new container version propagates to cvmfs) |
…els (with fallback for existing environments)
525f670 to
6af724c
Compare
ca3331a to
29eeddf
Compare
29eeddf to
427a8df
Compare
|
From the CI, numerical results are 1:1 identical. There are some minor aesthetic changes to the plots probably related to updated matplotlib, etc. The only substantive issue I can see so far is that "Preliminary" has disappeared from the plots, probably due to the second line of text getting lost somehow. Please others have a look as well though. https://cmsmwbot.web.cern.ch/WMassAnalysis/PRValidation/PR655/2026_02_07/ |
|
Ok the text subtitle issue is fixed. In principle this should be good to go now, but before it's merged first WMass/wums#15 needs to be merged and also we should figure out why the propagation of the updated container to cvmfs beyond CERN seems to be stuck/slow. |
|
Ok the propagation of CVMFS seems to be ok now (and the was merged) so this one can be merged now unless someone spots any other issues. |
Major upgrade of singularity container to v46 (see https://gitlab.cern.ch/bendavid/cmswmassdocker/-/commits/v46?ref_type=tags)
Some small corresponding technical updates which are needed (migrate python tensorflow lite usage to new litert package)
Also fix a few array out of bounds accesses (off by one errors) which caused problems after the ROOT/cling/etc update (but were already bugs previously in principle)
Note that critically now we are back to a situation again where we cannot/should not import tensorflow before the event loop (but we no longer need to)
I will see if a reasonable solution can be found to this stripping off symbols, but it might also be ok like this. (if it happens by mistake narf will detect it and give a descriptive error rather than failing mysteriously/unpredictably)