updated mpi doptuna

deephyper · Feb 21, 2024 · 7b60adf · 7b60adf
1 parent d1a8349
commit 7b60adf
Showing 1 changed file with 26 additions and 2 deletions.
diff --git a/deephyper_benchmark/search/_mpi_doptuna.py b/deephyper_benchmark/search/_mpi_doptuna.py
@@ -21,6 +21,7 @@
 from deephyper.core.utils._timeout import terminate_on_timeout  # noqa: E402
 from deephyper.evaluator import RunningJob  # noqa: E402
 from deephyper.search import Search  # noqa: E402
+from deephyper.skopt.moo import non_dominated_set
 
 
 def optuna_suggest_from_hp(trial, cs_hp):
@@ -284,7 +285,7 @@ def objective_wrapper(trial):
                 constraints = []
                 for i, lbi in enumerate(self._moo_lower_bounds):
                     if lbi is not None and type(output["objective"][i]) is not str:
-                        ci = -(output["objective"][i] - lbi) # <= 0
+                        ci = -(output["objective"][i] - lbi)  # <= 0
                         constraints.append(ci)
                 trial.set_user_attr("constraints", tuple(constraints))
 
@@ -356,8 +357,31 @@ def optimize_wrapper(duration):
 
             df_results = pd.DataFrame([t.user_attrs["results"] for t in all_trials])
             df_path = os.path.join(self._log_dir, "results.csv")
+
+            # Check if Multi-Objective Optimization was performed to save the pareto front
+            objective_columns = [
+                col for col in df_results.columns if col.startswith("objective")
+            ]
+
+            if len(objective_columns) > 1:
+                if pd.api.types.is_string_dtype(df_results[objective_columns[0]]):
+                    mask_no_failures = ~df_results[objective_columns[0]].str.startswith(
+                        "F"
+                    )
+                else:
+                    mask_no_failures = np.ones(len(df_results), dtype=bool)
+                objectives = -df_results.loc[
+                    mask_no_failures, objective_columns
+                ].values.astype(float)
+                mask_pareto_front = non_dominated_set(objectives)
+                df_results["pareto_efficient"] = False
+                df_results.loc[mask_no_failures, "pareto_efficient"] = mask_pareto_front
+
             df_results.to_csv(df_path, index=False)
 
-        self.extend_results_with_pareto_efficient(df_path)
+        if self.comm:
+            self.comm.Barrier()
+
         df_results = pd.read_csv(df_path)
+
         return df_results