SearchLibrium 0.0.149__tar.gz → 0.0.151__tar.gz

{searchlibrium-0.0.149/src/SearchLibrium.egg-info → searchlibrium-0.0.151}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: SearchLibrium
-Version: 0.0.149
+Version: 0.0.151
 Summary: A Python package for econometric models driven by search
 Author: Mario Fernando Taco Morales, Alexander Paz, Prithvi Beeramole, Robert Burdett
 Author-email: Zeke Ahern <z.ahern@qut.edu.au>

{searchlibrium-0.0.149 → searchlibrium-0.0.151}/pyproject.toml

@@ -82,7 +82,7 @@ Documentation = "https://pypi.org/project/SearchLibrium/"
 realpython = "SearchLibrium.__main__:main"
 
 [tool.bumpver]
-current_version = "0.0.149"
+current_version = "0.0.151"
 version_pattern = "MAJOR.MINOR.PATCH"
 commit_message = "[skip ci] Bump version {old_version} -> {new_version}"
 commit = true

{searchlibrium-0.0.149 → searchlibrium-0.0.151}/src/SearchLibrium/mdcev.py

@@ -353,6 +353,22 @@ class MDCEVModel:
 
         self.mle_success_ = bool(res.success)
         self.mle_message_ = str(res.message)
+
+        # fit_mle() refines baseline_utility_/alpha_/gamma_ beyond the initial
+        # self.fit(...) call above, but fit_result_ (what summary() reads)
+        # was never refreshed -- summary() would silently keep reporting the
+        # pre-MLE heuristic estimates. participation_rate/mean_allocation/
+        # mean_budget are observed-data statistics, unaffected by refinement.
+        old_result = self.fit_result_
+        self.fit_result_ = MDCEVFitResult(
+            labels=old_result.labels,
+            baseline_utility=self.baseline_utility_.copy(),
+            alpha=self.alpha_.copy(),
+            gamma=self.gamma_.copy(),
+            participation_rate=old_result.participation_rate,
+            mean_allocation=old_result.mean_allocation,
+            mean_budget=old_result.mean_budget,
+        )
         return self
 
     def summary(self) -> pd.DataFrame:

{searchlibrium-0.0.149 → searchlibrium-0.0.151}/src/SearchLibrium/predict.py

@@ -88,6 +88,14 @@ class DestinationPredictor:
         if prob is None:
             prob = getattr(self.model, 'choice_pred_prob', None)
         if prob is not None and hasattr(prob, 'shape') and len(prob.shape) == 2:
+            # JAX-fitted models (_jax=True, the default) leave this as a
+            # jax.Array. Downstream code (predict_destinations()) runs plain
+            # numpy ops (argmax/argsort) per row in a Python loop; numpy
+            # dispatching those against an un-converted jax.Array is
+            # catastrophically slow (~4s per argsort call observed, i.e.
+            # hours instead of seconds for a few thousand trips) rather than
+            # erroring, so convert once here up front.
+            prob = np.asarray(prob)
             self._probs = prob
             return prob
 
@@ -106,8 +114,9 @@ class DestinationPredictor:
             avail.reshape(self.n_trips, self.n_dests),
             return_logsum=True,
         )
+        probs = np.asarray(probs)  # see conversion note above
         self._probs = probs
-        self._logsums = logsums
+        self._logsums = np.asarray(logsums)
         return probs
 
     # ------------------------------------------------------------------
@@ -147,7 +156,11 @@ class DestinationPredictor:
                 (self.idca[self.trip_id_col] == tid)
                 & (self.idca[self.choice_col] == 1)
             ]
-            actual_code = int(actual[self.dest_col].iloc[0]) if len(actual) else -1
+            # dest_col isn't always integer-coded (e.g. string alt labels
+            # like 'CAR'/'SM'/'TRAIN') -- keep the raw value so it compares
+            # correctly against dest_codes/best_code, which retain whatever
+            # dtype the source data used.
+            actual_code = actual[self.dest_col].iloc[0] if len(actual) else -1
 
             rows.append({
                 self.trip_id_col: tid,

searchlibrium-0.0.151/src/SearchLibrium/version.txt

@@ -0,0 +1 @@
+0.0.151

{searchlibrium-0.0.149 → searchlibrium-0.0.151/src/SearchLibrium.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: SearchLibrium
-Version: 0.0.149
+Version: 0.0.151
 Summary: A Python package for econometric models driven by search
 Author: Mario Fernando Taco Morales, Alexander Paz, Prithvi Beeramole, Robert Burdett
 Author-email: Zeke Ahern <z.ahern@qut.edu.au>

searchlibrium-0.0.149/src/SearchLibrium/version.txt

@@ -1 +0,0 @@
-0.0.149