Attendees: Zenny, Jin, Sapta, Stefan

Apologies: Andrea, Olivier

PhD work:

• Zenny will concentrate this week on the producing a deployable version of the re-weighting infrastructure
• As of next week, also when Marco Za is back, will get back on the NLO work 

CMS integration:

• Jin presented slides (see attached)
• The cloning of the mg4gpu repo takes a long time
  
  • This should be fixed with the upcoming release. Only the "plugin" part will need to be cloned. Currently also all test processes, abstraction layers etc are being clone in additon. 
• Two problems have been spotted --> see slides
  
  • • use of "use_syst" 
    • pattern for G* also includes now newly created files starting with "Gpu" 
  • --> Olivier contacted
• lxplus8-gpu seems to be gone --> what is the replacement
  
  • lxplus-gpu is the new cluster name but one may end up on nodes with T4 GPUs, better to logon directly to lxplus801 which seems to be the only node with an A100 attached
• lxplus-condor, e.g. for DY+4j the process stops after 3 days
  
  • now queue could be found with a longer availability and GPUs attached --> SR to followup 
• for multi-threaded tar'ing "pigz" would be helpful on lxplus but is missing
  • --> SR will check
• In general the tar'ing for the gridpack takes too long. We can provide CMS a list (patterns?) of files which are not needed for the gridpack running and can be removed from the stage area before tar'ing
  • --> SR, Olivier to followup