Scaling ML meeting

US/Central

Questions and answers for the HPO talk: 

(Salman) Scale of number of GPUs needed for AID2E? Probably less than 1000

(Paolo) Do you need Panda to be the backen? No, it could be MPI

(Paolo) how can CCE help? 

(Lino) should PanDa/iDDS be one of the HPO frameworks 

(Walter) does this replace existing HPO frameworks, like raytune and Horovod? It can be integrated with other frameworks.

There are minutes attached to this event. Show them.