Workflows

ActivitySim workflows use the pypyr library. This task runner is more flexible than orca, and relies on isolated ‘context’ namespaces rather than storing tables and configurations in Python’s global state. Workflows are activated using the activitysim workflow command line interface.

You can run a workflow from a local working directory, or one of a number of pre-packaged workflows that are included with ActivitySim.

Workflows for Performance Monitoring

A collection of workflows used to compare the new sharrow code against legacy implementations can be found in the sharrow-contrast workflow subdirectory. Each of these first runs the relevant example model in test mode to compile the relevant functions, and then runs in production mode to measure runtime and memory usage. This is followed by another run in “legacy” mode that does not use sharrow, as well as a run in a reference implementation (version 1.1.3). Finally, a report is generated comparing the results of these various runs, to illustrate performance and validiate output quality.

Mini Examples

A collection of “mini” example workflows is provided for typical performance analysis. Each mini example is based on a small or medium size subarea for each model, and each runs in single-process mode to illustrate the performance and memory usage succinctly. These mini examples are sized so they should run on a typical modeler’s laptop (e.g. a modern CPU, 32 GB of RAM).

Workflow	Description
mtc_mini	Prototype MTC model on 50K households in San Francisco (190 zones)
arc_mini	Prototype ARC model on 10K households in Fulton County (1,296 zones)
sandag_1zone_mini	Placeholder SANDAG 1-Zone model on 50K households in a test region (90 zones)
sandag_2zone_mini	Placeholder SANDAG 2-Zone model on 50K households in a test region (90 TAZs, 690 MAZs)
sandag_3zone_mini	Placeholder SANDAG 3-Zone model on 50K households in a test region (90 TAZs, 690 MAZs)
sandag_xborder_mini	Prototype SANDAG Cross-Border model on 50K tours in a test region (94 TAZs, 586 MAZs)
psrc_mini	Placeholder PSRC Model on 50K households in Seattle (8400 MAZs)
comprehensive-mini	Runs all “mini” workflows listed above (warning, this takes a long time!)

To invoke a pre-packaged workflow, you can provide the workflow’s name and location within the activitysim/workflows directory. For example, to run the workflow of all “mini” examples, run:

activitysim workflow sharrow-contrast/comprehensive-mini

Full Size Examples

Running a “full size” example of an ActivitySim model generally requires more compute resources than the smaller performance tests shown above. These models generally require chunking to run, and servers with substantial amounts of RAM.

The “mp” workflows listed below will run each listed example in multiprocess mode. They will attempt to automatically scale the number of processes and the chunk size to the available resources of the server where they are run, but some models require substantial resources and may not run correctly on inadequate hardware.

Workflow	Description
mtc_mp	Prototype MTC model
arc_mp	Prototype ARC model, 5,922 zones
sandag_1zone_mp	Placeholder SANDAG 1-Zone model, 1.2M households and 4,984 zones