-
Notifications
You must be signed in to change notification settings - Fork 113
Project Meeting 2025.03.06
rcopperman edited this page Mar 6, 2025
·
3 revisions
- Technical
- Pandas 2.0 update
- PopulationSim scope
- Sub-Model Preprocessing scope
- Pandas 2.0 update
- Update from last year made software slower
- Spent last week researching why it is slower
- Run time for processing a lot of columns takes a lot of extra time causing long run times
- Next release of Pandas will have a fix for part of the slowdown, but not all of it
- Jeff is creating a script to band-aid (sub-optimal solution) the fix (sub-optimal solution), but may break if underlying data structures change
- Code may need to constantly updated
- Desirable change to move to PyArrow for better compatibility with Pandas, but it will be a high-effort change
- Documentation also requires use of Pandas 2.0
- Next step - spend a few hours to see if there is any better solution
- Population Sim
- Issue: Right now PopulationSim uses ActivitySim as dependency, but only version that works is 1.1.3 which is out of date
- Clarification: Purpose of task is create one install
- Do we keep them decouple them or make PopulationSim a dependency of ActivitySim; Should it be an optional install?
- If PopulationSim is a dependency of ActivitySim then it has to always remain consistent with ActivitySim
- Decision: Keep them separate
- Do we keep them decouple them or make PopulationSim a dependency of ActivitySim; Should it be an optional install?
- Clarification: Purpose of task is create one install
- Three tasks: Implement CI testing; Decouple PopulationSim from ActivitySim; Update package dependencies
- For these tasks, can move forward with making PopulationSim stand-up on its alone and then decision can be made later on about how it interacts with ActivitySim
- Issue: Right now PopulationSim uses ActivitySim as dependency, but only version that works is 1.1.3 which is out of date
- Sub-Model Preprocessing scope
- Not all models in ActivitySim have a preprocessor or postprocessor
- pre-processor is useful for calculations that are going to be repeated numerous times or are complex, clarity, and better efficiency in Sharrow
- Annotate post-processor is useful to derive and append attributes based on the model's output
- Lack of consistency between models may cause confusion and impede implementing some features
- Goal is to create a consistent framework that provides preprocessing and annotate postprocessing functionality to all sub-models in the ActivitySim code system
- First step: List of all sub-models and note whether they have a preprocessor and annotate postprocessor
- Not all models in ActivitySim have a preprocessor or postprocessor
- Tuesday's meeting is cancelled