Fix errors

Gurobi · Jan 2, 2025 · 2e03399 · 2e03399
1 parent 8893a6e
commit 2e03399
Show file tree

Hide file tree

Showing 8 changed files with 31 additions and 48 deletions.
diff --git a/docs/examples/example2_student_admission.py b/docs/examples/example2_student_admission.py
@@ -82,7 +82,6 @@
 """
 
 import gurobipy as gp
-import gurobipy_pandas as gppd
 import numpy as np
 import pandas as pd
 from sklearn.linear_model import LogisticRegression

diff --git a/notebooks/adversarial/adversarial_logistic_regression.ipynb b/notebooks/adversarial/adversarial_logistic_regression.ipynb
@@ -34,8 +34,7 @@
     "\n",
     "from sklearn.datasets import fetch_openml\n",
     "from sklearn.linear_model import LogisticRegression\n",
-    "from sklearn.pipeline import make_pipeline\n",
-    "from sklearn.preprocessing import StandardScaler"
+    "from sklearn.pipeline import make_pipeline"
    ]
   },
   {
@@ -66,7 +65,8 @@
    "outputs": [],
    "source": [
     "import sys\n",
-    "sys.path.append('../../src')"
+    "\n",
+    "sys.path.append(\"../../src\")"
    ]
   },
   {
@@ -77,8 +77,7 @@
    },
    "outputs": [],
    "source": [
-    "from gurobi_ml import add_predictor_constr\n",
-    "from gurobi_ml.sklearn import logistic_regression"
+    "from gurobi_ml import add_predictor_constr"
    ]
   },
   {
@@ -98,7 +97,9 @@
     "# Flatten and scale the data\n",
     "\n",
     "X = X.astype(np.float64) / 255.0\n",
-    "clf = LogisticRegression(C=50.0 / 500, penalty=\"l1\", solver=\"saga\", tol=0.1, random_state=4)\n",
+    "clf = LogisticRegression(\n",
+    "    C=50.0 / 500, penalty=\"l1\", solver=\"saga\", tol=0.1, random_state=4\n",
+    ")\n",
     "pipeline = make_pipeline(clf)\n",
     "pipeline.fit(X, y)"
    ]
@@ -122,7 +123,7 @@
    },
    "outputs": [],
    "source": [
-    "ex_prob = clf.predict_proba(X[imageno:imageno+1, :])\n",
+    "ex_prob = clf.predict_proba(X[imageno : imageno + 1, :])\n",
     "sorted_labels = np.argsort(ex_prob)[0]\n",
     "right_label = sorted_labels[-1]\n",
     "wrong_label = sorted_labels[-2]"
@@ -138,7 +139,7 @@
    "source": [
     "image = X[imageno, :]\n",
     "plt.imshow(image.reshape((28, 28)), cmap=\"gray\")\n",
-    "label = clf.predict(image.reshape(1,-1))\n",
+    "label = clf.predict(image.reshape(1, -1))\n",
     "print(f\"Solution is classified as {label}\")"
    ]
   },
@@ -165,7 +166,7 @@
     "delta = 10\n",
     "\n",
     "x = m.addMVar(image.shape, lb=0.0, ub=1.0, name=\"x\")\n",
-    "y = m.addMVar(ex_prob.shape, lb=-float('inf'), name=\"y\")\n",
+    "y = m.addMVar(ex_prob.shape, lb=-float(\"inf\"), name=\"y\")\n",
     "\n",
     "abs_diff = m.addMVar(image.shape, lb=0, ub=1, name=\"abs_diff\")\n",
     "\n",
@@ -177,7 +178,9 @@
     "m.addConstr(abs_diff >= -x + image)\n",
     "m.addConstr(abs_diff.sum() <= delta)\n",
     "\n",
-    "pred_constr = add_predictor_constr(m, pipeline.steps[-1][1], x, y, epsilon=0, predict_function='predict_proba')\n",
+    "pred_constr = add_predictor_constr(\n",
+    "    m, pipeline.steps[-1][1], x, y, epsilon=0, predict_function=\"predict_proba\"\n",
+    ")\n",
     "\n",
     "pred_constr.print_stats()"
    ]
@@ -214,7 +217,7 @@
    "outputs": [],
    "source": [
     "plt.imshow(x.X.reshape((28, 28)), cmap=\"gray\")\n",
-    "label = pipeline.predict(x.X.reshape(1,-1))\n",
+    "label = pipeline.predict(x.X.reshape(1, -1))\n",
     "print(f\"Solution is classified as {label}\")"
    ]
   },

diff --git a/notebooks/adversarial/adversarial_softmax.ipynb b/notebooks/adversarial/adversarial_softmax.ipynb
@@ -33,8 +33,7 @@
     "\n",
     "from sklearn.datasets import fetch_openml\n",
     "from sklearn.neural_network import MLPClassifier\n",
-    "from sklearn.pipeline import make_pipeline\n",
-    "from sklearn.preprocessing import StandardScaler"
+    "from sklearn.pipeline import make_pipeline"
    ]
   },
   {
@@ -65,7 +64,8 @@
    "outputs": [],
    "source": [
     "import sys\n",
-    "sys.path.append('../../src')"
+    "\n",
+    "sys.path.append(\"../../src\")"
    ]
   },
   {
@@ -76,9 +76,7 @@
    },
    "outputs": [],
    "source": [
-    "from gurobi_ml import add_predictor_constr\n",
-    "from gurobi_ml.sklearn import mlpregressor\n",
-    "from gurobi_ml.modeling import softmax"
+    "from gurobi_ml import add_predictor_constr"
    ]
   },
   {
@@ -112,7 +110,7 @@
    "outputs": [],
    "source": [
     "X = X.astype(np.float64) / 255.0\n",
-    "clf = MLPClassifier([20,20])\n",
+    "clf = MLPClassifier([20, 20])\n",
     "pipeline = make_pipeline(clf)\n",
     "pipeline.fit(X, Y)"
    ]
@@ -136,7 +134,7 @@
    },
    "outputs": [],
    "source": [
-    "ex_prob = clf.predict_proba(X[imageno:imageno+1, :])\n",
+    "ex_prob = clf.predict_proba(X[imageno : imageno + 1, :])\n",
     "sorted_labels = np.argsort(ex_prob)[0]\n",
     "right_label = sorted_labels[-1]\n",
     "wrong_label = sorted_labels[-2]"
@@ -171,7 +169,7 @@
    "outputs": [],
    "source": [
     "plt.imshow(image.reshape((28, 28)), cmap=\"gray\")\n",
-    "label = clf.predict(image.reshape(1,-1))\n",
+    "label = clf.predict(image.reshape(1, -1))\n",
     "print(f\"Solution is classified as {label}\")"
    ]
   },
@@ -198,7 +196,7 @@
     "m.addConstr(abs_diff.sum() <= delta)\n",
     "\n",
     "# Gurobi ML magic\n",
-    "pred_constr = add_predictor_constr(m, clf, x, y, out_activation='softmax')\n",
+    "pred_constr = add_predictor_constr(m, clf, x, y, out_activation=\"softmax\")\n",
     "\n",
     "pred_constr.print_stats()"
    ]
@@ -211,7 +209,7 @@
    },
    "outputs": [],
    "source": [
-    "m.Params.OBBT=3\n",
+    "m.Params.OBBT = 3\n",
     "m.optimize()"
    ]
   },
@@ -224,7 +222,7 @@
    "outputs": [],
    "source": [
     "plt.imshow(x.X.reshape((28, 28)), cmap=\"gray\")\n",
-    "label = pipeline.predict(x.X.reshape(1,-1))\n",
+    "label = pipeline.predict(x.X.reshape(1, -1))\n",
     "print(f\"Solution is classified as {label}\")"
    ]
   },

diff --git a/src/gurobi_ml/keras/keras.py b/src/gurobi_ml/keras/keras.py
@@ -112,7 +112,7 @@ def _mip_model(self, **kwargs):
                 pass
             elif isinstance(step, (keras.layers.ReLU, keras.layers.Softmax)):
                 layer = self._add_activation_layer(
-                    _input, self.act_dict["relu"](), output, name=f"relu{i}", **kwargs
+                    _input, self.act_dict["relu"], output, name=f"relu{i}", **kwargs
                 )
                 _input = layer.output
             else:
@@ -127,7 +127,7 @@ def _mip_model(self, **kwargs):
                     _input,
                     weights,
                     bias,
-                    self.act_dict[activation](),
+                    self.act_dict[activation],
                     output,
                     name=f"dense{i}",
                     **kwargs,

diff --git a/src/gurobi_ml/modeling/softmax.py b/src/gurobi_ml/modeling/softmax.py
@@ -165,10 +165,6 @@ def softmax(
     """Add the prediction constraints to Gurobi."""
     gp_model: gp.Model = predictor_model.gp_model
     output: gp.MVar = predictor_model.output
-    try:
-        predict_function: str = predictor_model.predict_function
-    except AttributeError:
-        predict_function = "predict_proba"
 
     if "epsilon" in kwargs:
         epsilon = kwargs["epsilon"]
@@ -190,7 +186,7 @@ def softmax(
 
         # Voila!
         gp_model.addConstr(
-            output == exponentials / denominator[:, np.newaxis], name=f"multlog"
+            output == exponentials / denominator[:, np.newaxis], name="multlog"
         )
     else:
         # How boy that is tedious you don't want not to use Gurobi 12!

diff --git a/src/gurobi_ml/sklearn/mlpregressor.py b/src/gurobi_ml/sklearn/mlpregressor.py
@@ -16,6 +16,7 @@
 """Module for formulating a :external+sklearn:py:class:`sklearn.neural_network.MLPRegressor` in a
 :external+gurobi:py:class:`Model`.
 """
+
 from ..exceptions import NoModel
 from ..modeling.neuralnet import BaseNNConstr
 from .skgetter import SKClassifier, SKRegressor
@@ -149,7 +150,7 @@ def _mip_model(self, **kwargs):
                 neural_net,
                 f"No implementation for activation function {neural_net.activation}",
             )
-        activation = self.act_dict[neural_net.activation]()
+        activation = self.act_dict[neural_net.activation]
 
         input_vars = self._input
         output = None
@@ -160,7 +161,7 @@ def _mip_model(self, **kwargs):
 
             # For last layer change activation
             if i == neural_net.n_layers_ - 2:
-                activation = self.act_dict[neural_net.out_activation_]()
+                activation = self.act_dict[neural_net.out_activation_]
                 output = self._output
                 if neural_net.out_activation_ in ("softmax", "logistic"):
                     kwargs["predict_function"] = self.predict_function

diff --git a/src/gurobi_ml/torch/sequential.py b/src/gurobi_ml/torch/sequential.py
@@ -127,7 +127,7 @@ def _mip_model(self, **kwargs):
                 activation = self.activations[type(step)]
                 layer = self._add_activation_layer(
                     _input,
-                    self.act_dict[activation](),
+                    self.act_dict[activation],
                     output,
                     name=f"{activation}_{i}",
                     **kwargs,

diff --git a/tests/test_sklearn/test_sklearn_formulations.py b/tests/test_sklearn/test_sklearn_formulations.py
@@ -70,11 +70,7 @@ def test_iris_proba(self):
         data = datasets.load_iris()
 
         X = data.data
-        y = data.target
 
-        # Make it a simple classification
-        X = X[y != 2]
-        y = y[y != 2]
         cases = IrisBinaryCases()
 
         for regressor in cases:
@@ -92,8 +88,6 @@ def test_iris_multi(self):
         data = datasets.load_iris()
 
         X = data.data
-        y = data.target
-
         # Make it a simple classification
         cases = IrisMultiCases()
 
@@ -106,11 +100,7 @@ def test_iris_clf(self):
         data = datasets.load_iris()
 
         X = data.data
-        y = data.target
 
-        # Make it a simple classification
-        X = X[y != 2]
-        y = y[y != 2]
         cases = IrisBinaryCases()
 
         for regressor in cases:
@@ -124,11 +114,7 @@ def test_iris_pwl_args(self):
         data = datasets.load_iris()
 
         X = data.data
-        y = data.target
 
-        # Make it a simple classification
-        X = X[y != 2]
-        y = y[y != 2]
         cases = IrisBinaryCases()
 
         for regressor in cases: