Agenda

• Admin - Phase 10A tasks

General Admin

• Phase 10A tasks
  • Consultants and Agencies priorities don't align according to spreadsheet
  • Probably want to allocate no more than half of resources for Phase 10 to these bundle of resources
  • Econometrically consistent simulation of multinomial and nested logit choices
    • Agree that it is a high priority
    • Question is true level of effort and level of risk
    • David H & Jeff N will reach out to Jan Zill to discuss level of effort
    • Need a design of the solution before implementation of solution
    • Agree Phase 10 funds should focus on design of the solution
  • PopSim tasks
    • PopSim needs focused effort
    • David H added three tasks
      • One task is to get PopSim to run
      • Should remove dependency of ActivitySim on PopSim
        • Alex's B dream scenario: One install w/ tools needed to run ActivitySim and PopSim; but appears that was never implemented in that way
        • Bo W and Alex B to expand on issue; add text
        • Include on short list for Phase 10A
      • Other tasks are related to performance improvements
        • Agreement that speeding up PopSim is a nice to have but not most important
        • Will not include on short list for Phase 10A
  • Add Global option to skip households in simulation when failed
    • Happens w/ edge cases or alternatives are not available
    • Problems are usually indicative of something being wrong and arguably you should fix what is wrong rather than "skip" the issue
    • Worthy of consideration, but may not make cut for Phase 10
  • Sharrow cache not being written to directory specified in sharrow_cache_dir
    • Fix that was shelved due to lack of funds and priorities
    • Jeff N does not know level of effort to fix it
    • Would be good to move to Pandas 2.0 because that is latest environment
    • Joe C says that Jeff N can spend existing Phase 9b funds on this task
    • New task added: Compatibility with Pandas 2.0
  • Unpredictable and sporadic memory release
    • This is a big unknown and unclear what to do about it
    • Jeff N does not know how to solve this problem
    • Will not pursue for Phase 10A
  • Validate all input files for components before running anything
    • Can open .yaml files and have tools necessary to open and check that everything is in place before uploading any data
    • Effort: Low risk (builds upon previous work) and low effort; flagged as medium effort by agencies
    • Not super critical for running models; but supports usability during model development; SANDAG says they do change up the config files relatively often
    • Propose keeping it on list for Phase 10A
  • Next steps
    • Like to take consensus approach to figuring out level of effort and assigning who will take on each task
    • Would be good to be able to set up funding such that consultants can knock off easy tasks if funds and time are available