Common settings for fastmath, sharrow_skip, and other compute control…

…s across components (#824)

* prevent crash when sharrow-test fails in interaction-simulate

* sharrow_fastmath

* activate sharrow without fastmath in example

* change fastmath default to F for escort

* skip and fastmath in sharrow_settings

* fix deprecation handling

* fix fastmath setting on escort configs

also forbid extra in pydantic settings

* pin pandera on other envs

* remove unused LOGSUM_CHOOSER_COLUMNS

* fix sharrow skip in simple simulate

* fix regressions

* sharrow_settings in cdap

* add missing sharrow_settings

* pandas compute options

* sharrow_settings -> compute_settings

* fix transit-pass-subsidy model for compute_settings

* strip outdated code

* change class name to PydanticCompute for clarity

activitysim/abm/models/atwork_subtour_frequency.py

@@ -117,6 +117,7 @@ def atwork_subtour_frequency(
         trace_label=trace_label,
         trace_choice_name="atwork_subtour_frequency",
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     # convert indexes to alternative names

activitysim/abm/models/atwork_subtour_scheduling.py

@@ -56,7 +56,6 @@ def atwork_subtour_scheduling(
     estimator = estimation.manager.begin_estimation(state, "atwork_subtour_scheduling")
 
     model_spec = state.filesystem.read_model_spec(file_name=model_settings.SPEC)
-    sharrow_skip = model_settings.sharrow_skip
     coefficients_df = state.filesystem.read_model_coefficients(model_settings)
     model_spec = simulate.eval_coefficients(
         state, model_spec, coefficients_df, estimator
@@ -96,7 +95,7 @@ def atwork_subtour_scheduling(
         estimator=estimator,
         chunk_size=state.settings.chunk_size,
         trace_label=trace_label,
-        sharrow_skip=sharrow_skip,
+        compute_settings=model_settings.compute_settings,
     )
 
     if estimator:

activitysim/abm/models/auto_ownership.py

@@ -9,20 +9,21 @@
 
 from activitysim.core import (
     config,
-    expressions,
     estimation,
+    expressions,
     simulate,
     tracing,
     workflow,
 )
 from activitysim.core.configuration.base import PreprocessorSettings, PydanticReadable
 from activitysim.core.configuration.logit import LogitComponentSettings
+
 from .util import annotate
 
 logger = logging.getLogger(__name__)
 
 
-class AutoOwnershipSettings(LogitComponentSettings):
+class AutoOwnershipSettings(LogitComponentSettings, extra="forbid"):
     """
     Settings for the `auto_ownership` component.
     """
@@ -99,6 +100,7 @@ def auto_ownership_simulate(
         trace_choice_name="auto_ownership",
         log_alt_losers=log_alt_losers,
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     if estimator:

activitysim/abm/models/cdap.py

@@ -17,7 +17,11 @@
     tracing,
     workflow,
 )
-from activitysim.core.configuration.base import PreprocessorSettings, PydanticReadable
+from activitysim.core.configuration.base import (
+    ComputeSettings,
+    PreprocessorSettings,
+    PydanticReadable,
+)
 from activitysim.core.util import reindex
 
 logger = logging.getLogger(__name__)
@@ -34,6 +38,7 @@ class CdapSettings(PydanticReadable, extra="forbid"):
     annotate_households: PreprocessorSettings | None = None
     COEFFICIENTS: Path
     CONSTANTS: dict[str, Any] = {}
+    compute_settings: ComputeSettings | None = None
 
 
 @workflow.step
@@ -202,6 +207,7 @@ def cdap_simulate(
             trace_hh_id=trace_hh_id,
             trace_label=trace_label,
             add_joint_tour_utility=add_joint_tour_utility,
+            compute_settings=model_settings.compute_settings,
         )
     else:
         choices = cdap.run_cdap(
@@ -215,6 +221,7 @@ def cdap_simulate(
             chunk_size=state.settings.chunk_size,
             trace_hh_id=trace_hh_id,
             trace_label=trace_label,
+            compute_settings=model_settings.compute_settings,
         )
 
     if estimator:

activitysim/abm/models/free_parking.py

@@ -118,6 +118,7 @@ def free_parking(
         trace_label=trace_label,
         trace_choice_name="free_parking_at_work",
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     free_parking_alt = model_settings.FREE_PARKING_ALT

activitysim/abm/models/joint_tour_composition.py

@@ -123,6 +123,7 @@ def joint_tour_composition(
         trace_label=trace_label,
         trace_choice_name="composition",
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     # convert indexes to alternative names

activitysim/abm/models/joint_tour_frequency.py

@@ -112,6 +112,7 @@ def joint_tour_frequency(
         trace_label=trace_label,
         trace_choice_name="joint_tour_frequency",
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     # convert indexes to alternative names

activitysim/abm/models/joint_tour_frequency_composition.py

@@ -26,7 +26,7 @@
 logger = logging.getLogger(__name__)
 
 
-class JointTourFrequencyCompositionSettings(LogitComponentSettings):
+class JointTourFrequencyCompositionSettings(LogitComponentSettings, extra="forbid"):
     """
     Settings for the `joint_tour_frequency_composition` component.
     """
@@ -140,6 +140,7 @@ def joint_tour_frequency_composition(
         trace_choice_name=trace_label,
         estimator=estimator,
         explicit_chunk_size=0,
+        compute_settings=model_settings.compute_settings,
     )
 
     if estimator:

activitysim/abm/models/joint_tour_participation.py

@@ -418,6 +418,7 @@ def joint_tour_participation(
         trace_choice_name="participation",
         custom_chooser=participants_chooser,
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     # choice is boolean (participate or not)

activitysim/abm/models/joint_tour_scheduling.py

@@ -104,7 +104,6 @@ def joint_tour_scheduling(
     estimator = estimation.manager.begin_estimation(state, "joint_tour_scheduling")
 
     model_spec = state.filesystem.read_model_spec(file_name=model_settings.SPEC)
-    sharrow_skip = model_settings.sharrow_skip
     coefficients_df = state.filesystem.read_model_coefficients(model_settings)
     model_spec = simulate.eval_coefficients(
         state, model_spec, coefficients_df, estimator
@@ -128,7 +127,7 @@ def joint_tour_scheduling(
         estimator=estimator,
         chunk_size=state.settings.chunk_size,
         trace_label=trace_label,
-        sharrow_skip=sharrow_skip,
+        compute_settings=model_settings.compute_settings,
     )
 
     if estimator:

activitysim/abm/models/location_choice.py

@@ -192,6 +192,9 @@ def _location_sample(
         chunk_tag=chunk_tag,
         trace_label=trace_label,
         zone_layer=zone_layer,
+        compute_settings=model_settings.compute_settings.subcomponent_settings(
+            "sample"
+        ),
     )
 
     return choices
@@ -696,6 +699,9 @@ def run_location_simulate(
         trace_choice_name=model_settings.DEST_CHOICE_COLUMN_NAME,
         estimator=estimator,
         skip_choice=skip_choice,
+        compute_settings=model_settings.compute_settings.subcomponent_settings(
+            "simulate"
+        ),
     )
 
     if not want_logsums:

activitysim/abm/models/mandatory_tour_frequency.py

@@ -53,7 +53,7 @@ def add_null_results(state, trace_label, mandatory_tour_frequency_settings):
     state.add_table("persons", persons)
 
 
-class MandatoryTourFrequencySettings(LogitComponentSettings):
+class MandatoryTourFrequencySettings(LogitComponentSettings, extra="forbid"):
     """
     Settings for the `mandatory_tour_frequency` component.
     """
@@ -134,6 +134,7 @@ def mandatory_tour_frequency(
         trace_label=trace_label,
         trace_choice_name="mandatory_tour_frequency",
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     # convert indexes to alternative names

activitysim/abm/models/non_mandatory_tour_frequency.py

@@ -12,8 +12,8 @@
 
 from activitysim.abm.models.util import annotate
 from activitysim.abm.models.util.overlap import (
-    person_max_window,
     person_available_periods,
+    person_max_window,
 )
 from activitysim.abm.models.util.school_escort_tours_trips import (
     recompute_tour_count_statistics,
@@ -161,7 +161,7 @@ class NonMandatoryTourSpecSegment(PydanticReadable):
     COEFFICIENTS: Path
 
 
-class NonMandatoryTourFrequencySettings(LogitComponentSettings):
+class NonMandatoryTourFrequencySettings(LogitComponentSettings, extra="forbid"):
     """
     Settings for the `non_mandatory_tour_frequency` component.
     """
@@ -321,6 +321,7 @@ def non_mandatory_tour_frequency(
             trace_choice_name="non_mandatory_tour_frequency",
             estimator=estimator,
             explicit_chunk_size=model_settings.explicit_chunk,
+            compute_settings=model_settings.compute_settings,
         )
 
         if estimator:

activitysim/abm/models/school_escorting.py

@@ -3,7 +3,7 @@
 from __future__ import annotations
 
 import logging
-from typing import Any
+from typing import Any, Literal
 
 import numpy as np
 import pandas as pd
@@ -367,7 +367,7 @@ def create_school_escorting_bundles_table(choosers, tours, stage):
     return bundles
 
 
-class SchoolEscortSettings(BaseLogitComponentSettings):
+class SchoolEscortSettings(BaseLogitComponentSettings, extra="forbid"):
     """
     Settings for the `telecommute_frequency` component.
     """
@@ -391,21 +391,6 @@ class SchoolEscortSettings(BaseLogitComponentSettings):
     GENDER_WEIGHT: float = 10.0
     AGE_WEIGHT: float = 1.0
 
-    sharrow_skip: bool | dict[str, bool] = False
-    """Setting to skip sharrow.
-
-    Sharrow can be skipped (or not) for all school escorting stages by giving
-    simply true or false.  Alternatively, it can be skipped only for particular
-    stages by giving a mapping of stage name to skipping.  For example:
-
-    ```yaml
-    sharrow_skip:
-        OUTBOUND: true
-        INBOUND: false
-        OUTBOUND_COND: true
-    ```
-    """
-
     SIMULATE_CHOOSER_COLUMNS: list[str] | None = None
 
     SPEC: None = None
@@ -428,6 +413,13 @@ class SchoolEscortSettings(BaseLogitComponentSettings):
     explicit_chunk: int = 0
     """If > 0, use this chunk size instead of adaptive chunking."""
 
+    LOGIT_TYPE: Literal["MNL"] = "MNL"
+    """Logit model mathematical form.
+
+    * "MNL"
+        Multinomial logit model.
+    """
+
 
 @workflow.step
 def school_escorting(
@@ -508,19 +500,15 @@ def school_escorting(
             state, model_spec_raw, coefficients_df, estimator
         )
 
-        # allow for skipping sharrow entirely in this model with `sharrow_skip: true`
+        # allow for skipping sharrow entirely in this model with `compute_settings.sharrow_skip: true`
         # or skipping stages selectively with a mapping of the stages to skip
-        sharrow_skip = model_settings.sharrow_skip
-        stage_sharrow_skip = False  # default is false unless set below
-        if sharrow_skip:
-            if isinstance(sharrow_skip, dict):
-                stage_sharrow_skip = sharrow_skip.get(stage.upper(), False)
-            else:
-                stage_sharrow_skip = True
-        if stage_sharrow_skip:
-            locals_dict["_sharrow_skip"] = True
-        else:
-            locals_dict.pop("_sharrow_skip", None)
+        stage_compute_settings = model_settings.compute_settings.subcomponent_settings(
+            stage.upper()
+        )
+        # if stage_sharrow_skip:
+        #     locals_dict["_sharrow_skip"] = True
+        # else:
+        #     locals_dict.pop("_sharrow_skip", None)
 
         # reduce memory by limiting columns if selected columns are supplied
         chooser_columns = model_settings.SIMULATE_CHOOSER_COLUMNS
@@ -584,6 +572,7 @@ def school_escorting(
             trace_choice_name="school_escorting_" + stage,
             estimator=estimator,
             explicit_chunk_size=model_settings.explicit_chunk,
+            compute_settings=stage_compute_settings,
         )
 
         if estimator:

activitysim/abm/models/stop_frequency.py

@@ -214,6 +214,7 @@ def stop_frequency(
             trace_label=tracing.extend_trace_label(trace_label, segment_name),
             trace_choice_name="stops",
             estimator=estimator,
+            compute_settings=model_settings.compute_settings,
         )
 
         # convert indexes to alternative names

activitysim/abm/models/telecommute_frequency.py

@@ -20,7 +20,7 @@
 logger = logging.getLogger("activitysim")
 
 
-class TelecommuteFrequencySettings(LogitComponentSettings):
+class TelecommuteFrequencySettings(LogitComponentSettings, extra="forbid"):
     """
     Settings for the `telecommute_frequency` component.
     """
@@ -99,6 +99,7 @@ def telecommute_frequency(
         trace_label=trace_label,
         trace_choice_name="telecommute_frequency",
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     choices = pd.Series(model_spec.columns[choices.values], index=choices.index)

activitysim/abm/models/transit_pass_ownership.py

@@ -20,7 +20,7 @@
 logger = logging.getLogger("activitysim")
 
 
-class TransitPassOwnershipSettings(LogitComponentSettings):
+class TransitPassOwnershipSettings(LogitComponentSettings, extra="forbid"):
     """
     Settings for the `transit_pass_ownership` component.
     """
@@ -93,6 +93,7 @@ def transit_pass_ownership(
         trace_label=trace_label,
         trace_choice_name="transit_pass_ownership",
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     if estimator:

activitysim/abm/models/transit_pass_subsidy.py

@@ -92,6 +92,7 @@ def transit_pass_subsidy(
         trace_label=trace_label,
         trace_choice_name="transit_pass_subsidy",
         estimator=estimator,
+        compute_settings=model_settings.compute_settings,
     )
 
     if estimator: