Skip to content

Performance Test Subarea Model #1078

Description

@jpn--

We would like to have a "sub-area" model to better evaluate runtime performance on the example two-zone model. The two zone example model is based on the SANDAG model.

The current "demo" size of this model, which is written in to the repository in the data directory, has 874 MAZs and 385 TAZs. Meanwhile the "full size" model has 4947 TAZs and 24333 MAZs.

We believe the current demo size model is too small to rigorously exercise and measure model performance, while the full size model is too large and unwieldy to use for regular testing. We would like a subarea model somewhere in between these two. Ideally, it should be large enough to stress test performance (runtime and memory usage) but small enough to run on a laptop with 32GB of RAM in under half an hour.

With this model, we are NOT interested in the statistical performance -- good calibrated results are unimportant. But the area and the synthetic population should be diverse enough to cover the modeling space -- households and behaviors of every type.

cc @ActivitySim/engineering @JoeJimFlood

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    FeatureNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions