Support explicit_chunk for mode-choice components

[vincentgong7]

Description

What

Plumb the existing explicit_chunk setting through to the mode-choice components
(tour_mode_choice, trip_mode_choice), matching how location, destination, and scheduling
components already use it.

Why

explicit_chunk (a fixed per-chunk chooser count, active under chunk_training_mode: explicit) lets a
component run with a deterministic, bounded memory footprint. Location/destination/scheduling
components honor it via LocationComponentSettings.explicit_chunk, but mode-choice components never
received it — the field didn't exist on their settings class and mode_choice_simulate /
simple_simulate neither accepted nor forwarded it.

On memory-constrained machines the adaptive chunker grows each chunk toward its memory budget, so peak
RSS is roughly constant regardless of sample size — which can OOM large models even after subsampling.
Fixed explicit_chunk sizes bound the peak. This change makes that strategy usable for mode choice too;
for us it was the last unchunkable core step when fitting a full-sample run (8.1M trips, 7787 zones) onto
a 64 GB node.

What changed (backward-compatible — default 0 = unchanged behavior)

• TemplatedLogitComponentSettings: new explicit_chunk: float = 0 field (the base for mode-choice
  settings; mirrors the field already on LocationComponentSettings).
• mode_choice_simulate(...): new explicit_chunk_size parameter, forwarded to simple_simulate.
• run_tour_mode_choice_simulate(...) and trip_mode_choice(...): pass
  explicit_chunk_size=model_settings.explicit_chunk.
• simple_simulate(...): new explicit_chunk_size parameter, forwarded to
  chunk.adaptive_chunked_choosers.

4 files changed, 16 insertions(+), 1 deletion(-).

Usage

# settings.yaml
chunk_training_mode: explicit
# tour_mode_choice.yaml / trip_mode_choice.yaml
explicit_chunk: 5000

Testing

Verified on a full-sample run (8.1M trips, 7787 zones) on a 64 GB node: tour/trip mode choice ran with a
bounded per-chunk memory footprint under explicit_chunk: 5000; with the default (0) behavior is
identical to before.

Acknowledgement

This work was supported by the XCarcity project. https://xcarcity.nl

[Copilot]

Pull request overview

This PR extends ActivitySim’s existing explicit_chunk chunking control to mode-choice components, allowing deterministic, bounded chooser chunk sizes during tour/trip mode choice simulation (via chunk_training_mode: explicit).

Changes:

• Add explicit_chunk to TemplatedLogitComponentSettings so mode-choice settings can declare fixed chunk sizing.
• Add/forward explicit_chunk_size through mode_choice_simulate(...) and simple_simulate(...) into chunk.adaptive_chunked_choosers(...).
• Wire model_settings.explicit_chunk into tour and trip mode choice simulation calls.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
activitysim/core/simulate.py	Adds explicit_chunk_size to simple_simulate and forwards it to the chooser chunking generator.
activitysim/core/configuration/logit.py	Adds explicit_chunk to TemplatedLogitComponentSettings so templated/segmented logit components (incl. mode choice) can configure fixed chunking.
activitysim/abm/models/util/mode.py	Threads explicit_chunk_size through mode_choice_simulate and passes model_settings.explicit_chunk from tour mode choice runner.
activitysim/abm/models/trip_mode_choice.py	Passes model_settings.explicit_chunk into mode_choice_simulate for trip mode choice.

Comments suppressed due to low confidence (1)

activitysim/abm/models/util/mode.py:43

• The mode_choice_simulate docstring parameter list is now out of sync with the function signature: it still mentions chunk_size (which is not a parameter) and doesn't document explicit_chunk_size or the updated optional type for compute_settings. This can mislead callers when enabling explicit chunking.

    explicit_chunk_size: float = 0,
):
    """
    common method for  both tour_mode_choice and trip_mode_choice

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.