Bugfix for trip scheduling choice (ActivitySim#884)

* remove problem code for sharrow tracing * do not allow zero probs in trip schedule choice * remove problem code for sharrow load --------- Co-authored-by: Sijia Wang <[email protected]>
RSGInc · Aug 9, 2024 · c6aa6d7 · c6aa6d7
1 parent 9181390
commit c6aa6d7
Show file tree

Hide file tree

Showing 2 changed files with 31 additions and 24 deletions.
diff --git a/activitysim/abm/models/trip_scheduling_choice.py b/activitysim/abm/models/trip_scheduling_choice.py
@@ -291,7 +291,7 @@ def run_trip_scheduling_choice(
                 alternatives=schedules,
                 spec=spec,
                 choice_column=SCHEDULE_ID,
-                allow_zero_probs=True,
+                allow_zero_probs=False,
                 zero_prob_choice_val=-999,
                 log_alt_losers=False,
                 want_logsums=False,

diff --git a/activitysim/core/simulate.py b/activitysim/core/simulate.py
@@ -722,19 +722,22 @@ def eval_utilities(
         offsets = np.nonzero(list(trace_targets))[0]
 
         # trace sharrow
-        if sh_flow is not None:
-            try:
-                data_sh = sh_flow.load(
-                    sh_tree.replace_datasets(
-                        df=choosers.iloc[offsets],
-                    ),
-                    dtype=np.float32,
-                )
-                expression_values_sh = pd.DataFrame(data=data_sh.T, index=spec.index)
-            except ValueError:
-                expression_values_sh = None
-        else:
-            expression_values_sh = None
+        # TODO: This block of code is sometimes extremely slow or hangs for no apparent
+        #       reason. It is temporarily disabled until the cause can be identified, so
+        #       that most tracing can be still be done with sharrow enabled.
+        # if sh_flow is not None:
+        #     try:
+        #         data_sh = sh_flow.load(
+        #             sh_tree.replace_datasets(
+        #                 df=choosers.iloc[offsets],
+        #             ),
+        #             dtype=np.float32,
+        #         )
+        #         expression_values_sh = pd.DataFrame(data=data_sh.T, index=spec.index)
+        #     except ValueError:
+        #         expression_values_sh = None
+        # else:
+        expression_values_sh = None
 
         # get array of expression_values
         # expression_values.shape = (len(spec), len(choosers))
@@ -806,16 +809,20 @@ def eval_utilities(
                 )
                 print(f"{sh_util.shape=}")
                 print(misses)
-                _sh_flow_load = sh_flow.load(sh_tree)
-                print("possible problematic expressions:")
-                for expr_n, expr in enumerate(exprs):
-                    closeness = np.isclose(
-                        _sh_flow_load[:, expr_n], expression_values[expr_n, :]
-                    )
-                    if not closeness.all():
-                        print(
-                            f"  {closeness.sum()/closeness.size:05.1%} [{expr_n:03d}] {expr}"
-                        )
+                # load sharrow flow
+                # TODO: This block of code is sometimes extremely slow or hangs for no apparent
+                #       reason. It is temporarily disabled until the cause can be identified, so
+                #       that model does not hang with sharrow enabled.
+                # _sh_flow_load = sh_flow.load(sh_tree)
+                # print("possible problematic expressions:")
+                # for expr_n, expr in enumerate(exprs):
+                #     closeness = np.isclose(
+                #         _sh_flow_load[:, expr_n], expression_values[expr_n, :]
+                #     )
+                #     if not closeness.all():
+                #         print(
+                #             f"  {closeness.sum()/closeness.size:05.1%} [{expr_n:03d}] {expr}"
+                #         )
                 raise
         except TypeError as err:
             print(err)