Project Meeting 2024.01.04 - ActivitySim/activitysim GitHub Wiki

Agenda

  • Close out Phase 8 tasks
  • Admin Items

Action Items

  • WSP to complete CS pull request for Component Documentation
  • Update on Data Type Optimization PR
  • CS to complete RSG's input checker pull request
  • Joe to draft scopes of work for Phase 9 general support and PM tasks

Meeting Notes

Close out Phase 8 tasks

  • CS
    • Project Mgmt - completed
    • Unit Testing: Replace Orca - completed
    • New Data Pipeline File Format- completed
    • Configuration Documentation & Input Checking
      • PR is being reviewed by Sijia
    • Component Documentation
      • PR is being reviewed by Sijia
    • User’s Guide
      • Complete but waiting to merge at the same time as Configuration Documentation & Input Checking PR
  • WSP
    • Roadmap - completed
      • Documentation has been delivered and billed out. The partners are taking the lead with any changes and pulling together something that is unified.
    • Data Type Update / Memory Profiling
      • Memory profiling – memo done and went through two rounds of edits.
      • Data Type Optimization - PR is open and includes background information/documentation
      • Memory leak PR was reviewed and merged.
  • RSG
    • Input Checker
      • Jeff made a couple of updates to fix some testing issues, to make them compatible with the input checker. Another day or two and code review will be done.

Admin Items: Phase 9 contracting

  • Main contracts are in place for all bench contracts. New task orders can be issued.
  • Due to complexities of received proposals for Phase 9, and high costs associated with tradeoffs, it is taking more time to prioritize. During those discussion, concerns were raised about previous performance enhancements deliverables that did not perform well for everyone. In particular, SANDAG and MWCOG experienced issues with implementing sharrow.
    • SANDAG gave up and moved back to a non-sharrow implementation.
    • MWCOG was similarly having performance issues and giving up on sharrow features. For MWCOG, sharrow struggles with workplace location choice and district-level constants. When you remove them, it runs much faster (not sure if that’s true without sharrow). Also, it recompiles randomly, not sure why that is. They are hoping to get Jeff’s help to figure these things out.
    • However, PSRC showed 40%+ run time improvement with Sharrow. They also did a thorough job to make sure expressions were written in a sharrow friendly way.
  • This process highlighted some communication issues. Consortium members were having issues and not communicating to the group. The consortium doesn't want to get in people’s business but it might have been good to have more collective conversations about issues in order to identify common problems that the consortium should be investing in. There should be a way to create shared visibility for common issues.
  • The proposed path forward includes:
    • Pick example models now (no objections to the following):
      • 1-zone MTC example (extended)
      • 2-zone BAYDAG example (MTC geographic region)
    • Go through performance issues with a fine tooth comb for these two examples and optimize ActivitySim 1, based on these two models
    • Proposal to fast track the estimation stuff and multiprocessing as well
    • Document all the examples and provide guidance for proper specification
    • Invest non-trivial amount of resources to complete this work
    • Do this work before we do anything else
  • Contracting
    • Two paths forward – quick task orders that include:
      • Small funds to keep people going while the group works through these concerns
      • Moderate funds to keep people going and do ActivitySim 1 optimizations
    • Phase 9 larger tasks to hold off until we have more information
    • Proposal to set aside funds for people to address issues found in specific implementations for something that another consultant developed
  • Overall takeaways of how to turn these issues into lessons learned and pathways forward:
    • Need to work towards making models more similar than not, to provide templates that work for everyone
    • Make sure that there robustness in skills, so multiple people can address issues more promptly
  • General agreement
    • Need to fast track optimization of examples
    • Joe to discuss with consultants + PSRC, SANDAG, MTC to get scope language
    • Joe will work on drafting scopes of general support/PM tasks