Include objective sense in benchmark file; update charts

benchmark/benchmark.py

@@ -114,6 +114,13 @@ def charts():
     benchmark.load_results("%s/benchmark_ml.csv" % basepath)
     results = benchmark.raw_results()
     results["Gap (%)"] = results["Gap"] * 100.0
+    
+    sense = results.loc[0, "Sense"]
+    if sense == "min":
+        primal_column = "Relative Upper Bound"
+    else:
+        primal_column = "Relative Lower Bound"
+        
     palette={
         "baseline": "#9b59b6", 
         "ml-exact": "#3498db",
@@ -151,7 +158,7 @@ def charts():
                  );
     axes[2].set_ylim(0.95,1.01)
     sns.stripplot(x="Solver",
-                  y="Relative Lower Bound",
+                  y=primal_column,
                   jitter=0.25,
                   data=results[results["Solver"] == "ml-heuristic"],
                   ax=axes[2],

miplearn/benchmark.py

@@ -31,7 +31,10 @@ class BenchmarkRunner:
                                             label="Solve (%s)" % name,
                                             collect_training_data=False)
             for i in range(len(instances)):
-                self._push_result(results[i], solver=solver, name=name, instance=i)
+                self._push_result(results[i],
+                                  solver=solver,
+                                  name=name,
+                                  instance=i)
     
     def raw_results(self):
         return self.results
@@ -60,6 +63,7 @@ class BenchmarkRunner:
                                                  "Gap",
                                                  "Nodes",
                                                  "Mode",
+                                                 "Sense",
                                                 ])
         lb = result["Lower bound"]
         ub = result["Upper bound"]
@@ -73,6 +77,7 @@ class BenchmarkRunner:
             "Gap": gap,
             "Nodes": result["Nodes"],
             "Mode": solver.mode,
+            "Sense": result["Sense"],
         }, ignore_index=True)
         groups = self.results.groupby("Instance")
         best_lower_bound = groups["Lower Bound"].transform("max")

miplearn/solvers.py

@@ -94,8 +94,13 @@ class InternalSolver:
                     (count_fixed, count_total))
                 
     def set_model(self, model):
+        from pyomo.core.kernel.objective import minimize, maximize
         self.model = model
         self.solver.set_instance(model)
+        if self.solver._objective.sense == minimize:
+            self.sense = "min"
+        else:
+            self.sense = "max"
         self.var_name_to_var = {}
         for var in model.component_objects(Var):
             self.var_name_to_var[var.name] = var
@@ -140,6 +145,7 @@ class GurobiSolver(InternalSolver):
             "Upper bound": results["Problem"][0]["Upper bound"],
             "Wallclock time": results["Solver"][0]["Wallclock time"],
             "Nodes": self.solver._solver_model.getAttr("NodeCount"),
+            "Sense": self.sense,
         }    
             
     def _load_vars(self):
@@ -179,6 +185,7 @@ class CPLEXSolver(InternalSolver):
             "Upper bound": results["Problem"][0]["Upper bound"],
             "Wallclock time": results["Solver"][0]["Wallclock time"],
             "Nodes": 1,
+            "Sense": self.sense,
         }
     
     def solve_lp(self, tee=False):