Please read these instructions before posting any event on Fermilab Indico

Indico will be unavailable on Wed., Dec 18th from 7 - 7:30am due to server maintenance.

Scaling ML meeting

US/Central

News

Frontera will be reserved for Texascale Days, very large scale runs of at least half system size (3500 - 4096 nodes) or full system size (7000 - 8192 nodes), during the second week of December after the Frontera Maintenance on December 10th. 

NSBI and Omnifold:

     Questions on the NSBI (WH), ask Aishik in Slack:

  • *the dataset size in EOS
  • * HEPSim framework to simulate data as a toy data for the scaling study
  • * What prevents the code from being public
  • * How many hours were spent in training the models and the rest of the pipeline

     Try the complex workflow tool (XY):

  • RADICAL: https://radical-cybertools.github.io/radical-pilot/index.html and tutorials: https://radicalpilot.readthedocs.io/en/stable/tutorials.html
  • RADICAL-Ensemble toolkits are probably more suitable for the NSBI study: https://radicalentk.readthedocs.io/en/stable/;
  • https://radicalentk.readthedocs.io/en/stable/examples/poe.html

    Ways of Copying data using Globus:1) One can use the US ATLAS endpoint if the data is for production; 2) Download the data to a HPC center (e.g. replicate the dataset to BNL) and then use Globus. Group space in BNL is limited to 500 GB. 

Inference as a Service (XY)

  • Deploy the Server in Chicago User Facility, or 
  • NRP from UC Dan Diego, which US ATLAS has allocations
There are minutes attached to this event. Show them.
    • 14:30 14:45
      Intro 15m
      Speakers: Paolo Calafiura (LBNL), Walter Hopkins (Argonne National Laboratory)
    • 14:45 15:00
      AOB 15m