Import keras directly

README.md

@@ -13,7 +13,7 @@
 
 Gurobi Machine Learning is an [open-source](https://gurobi-machinelearning.readthedocs.io/en/latest/meta/license.html) python package to formulate trained regression models in a [`gurobipy`](https://pypi.org/project/gurobipy/) model to be solved with the Gurobi solver.
 
-The package currently supports various [scikit-learn](https://scikit-learn.org/stable/) objects. It has limited support for the [Keras](https://keras.io/) API of [TensorFlow](https://www.tensorflow.org/), [PyTorch](https://pytorch.org/) and [XGBoost](https://www.xgboost.ai). Only neural networks with ReLU activation can be used with Keras and PyTorch.
+The package currently supports various [scikit-learn](https://scikit-learn.org/stable/) objects. It has limited support for [Keras](https://keras.io/), [PyTorch](https://pytorch.org/) and [XGBoost](https://www.xgboost.ai). Only neural networks with ReLU activation can be used with Keras and PyTorch.
 
 # Documentation
 

docs/source/conf.py

@@ -81,7 +81,7 @@ def get_versions(file: Path):
 .. |PandasVersion| replace:: {dep_versions["pandas"]}
 .. |TorchVersion| replace:: {dep_versions["torch"]}
 .. |SklearnVersion| replace:: {dep_versions["scikit-learn"]}
-.. |TensorflowVersion| replace:: {dep_versions["tensorflow"]}
+.. |KerasVersion| replace:: {dep_versions["keras"]}
 .. |XGBoostVersion| replace:: {dep_versions["xgboost"]}
 .. |LightGBMVersion| replace:: {dep_versions["lightgbm"]}
 .. |VariablesDimensionsWarn| replace:: {VARS_SHAPE}
@@ -162,7 +162,7 @@ def get_versions(file: Path):
 # so a file named "default.css" will overwrite the builtin "default.css".
 html_static_path = ["_static"]
 autodoc_member_order = "groupwise"
-autodoc_mock_imports = ["torch", "tensorflow", "xgboost"]
+autodoc_mock_imports = ["torch", "keras", "tensorflow", "xgboost"]
 html_css_files = [
     "gurobi_ml.css",
 ]

docs/source/user/start.rst

@@ -29,9 +29,8 @@ The package currently supports various `scikit-learn
 <https://scikit-learn.org/stable/>`_ objects. It can also formulate
 gradient boosting regression models from `XGboost <https://xgboost.readthedocs.io/en/stable/>`_
 and `LightGBM <https://lightgbm.readthedocs.io/en/stable/>`.
-Finally, it has limited support for the
-`Keras <https://keras.io/>`_ API of `TensorFlow <https://www.tensorflow.org/>`_
-and `PyTorch <https://pytorch.org/>`_. Only neural networks with ReLU activation
+Finally, it has limited support for
+`Keras <https://keras.io/>`_. Only neural networks with ReLU activation
 can be used with these two packages.
 
 The package is actively developed and users are encouraged to :doc:`contact us
@@ -87,16 +86,16 @@ We encourage to install the package via pip (or add it to your
        - |TorchVersion|
      * - :pypi:`scikit-learn`
        - |SklearnVersion|
-     * - :pypi:`tensorflow`
-       - |TensorflowVersion|
+     * - :pypi:`keras`
+       - |KerasVersion|
      * - :pypi:`xgboost`
        - |XGBoostVersion|
      * - :pypi:`lightgbm`
        - |LightGBMVersion|
 
   Installing any of the machine learning packages is only required if the
   predictor you want to insert uses them (i.e. to insert a Keras based predictor
-  you need to have :pypi:`tensorflow` installed).
+  you need to have :pypi:`keras` installed).
 
 
 Usage

notebooks/dev_requirements.txt

@@ -1,14 +1,9 @@
 ipywidgets
 matplotlib
-myst_nb
 notebook
 pandas
 seaborn
-Sphinx
-sphinx-copybutton
-sphinx-pyproject
-sphinx-rtd-theme
-tensorflow
+keras
 torch
 skorch
 ../.

requirements.keras.txt

@@ -1 +1,2 @@
 tensorflow==2.18.0
+keras==3.7.0

src/gurobi_ml/keras/keras.py

@@ -16,7 +16,7 @@
 """Module for formulating a Keras model into a :external+gurobi:py:class:`Model`."""
 
 import numpy as np
-from tensorflow import keras
+import keras
 
 from ..exceptions import NoModel, NoSolution
 from ..modeling.neuralnet import BaseNNConstr

src/gurobi_ml/registered_predictors.py

@@ -90,8 +90,8 @@ def lightgbm_convertors():
 
 def keras_convertors():
     """Collect known Keras objects that can be embedded and the conversion class."""
-    if "tensorflow" in sys.modules:
-        from tensorflow import keras  # pylint: disable=import-outside-toplevel
+    if "keras" in sys.modules:
+        import keras  # pylint: disable=import-outside-toplevel
 
         from .keras import add_keras_constr  # pylint: disable=import-outside-toplevel
 

tests/test_keras/test_keras_exceptions.py

@@ -2,8 +2,7 @@
 
 import gurobipy as gp
 import numpy as np
-import tensorflow as tf
-from tensorflow import keras
+import keras
 
 from gurobi_ml import add_predictor_constr
 from gurobi_ml.exceptions import NoModel
@@ -22,9 +21,9 @@ def setUp(self) -> None:
 
     def do_test(self, nn):
         nn.compile(
-            optimizer=tf.keras.optimizers.Adam(0.001),
-            loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
-            metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],
+            optimizer=keras.optimizers.Adam(0.001),
+            loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),
+            metrics=[keras.metrics.SparseCategoricalAccuracy()],
         )
 
         nn.fit(
@@ -49,12 +48,12 @@ def test_keras_bad_activation(self):
         self.x_train = np.reshape(self.x_train, (-1, 28 * 28))
         self.x_test = np.reshape(self.x_test, (-1, 28 * 28))
 
-        nn = tf.keras.models.Sequential(
+        nn = keras.models.Sequential(
             [
-                tf.keras.layers.InputLayer((28 * 28,)),
-                tf.keras.layers.Dense(50, activation="sigmoid"),
-                tf.keras.layers.Dense(50, activation="relu"),
-                tf.keras.layers.Dense(10),
+                keras.layers.InputLayer((28 * 28,)),
+                keras.layers.Dense(50, activation="sigmoid"),
+                keras.layers.Dense(50, activation="relu"),
+                keras.layers.Dense(10),
             ]
         )
         self.do_test(nn)
@@ -65,19 +64,19 @@ def test_keras_layers(self):
         self.x_train = np.reshape(self.x_train, (-1, 28, 28, 1))
         self.x_test = np.reshape(self.x_test, (-1, 28, 28, 1))
 
-        nn = tf.keras.models.Sequential(
+        nn = keras.models.Sequential(
             [
-                tf.keras.layers.InputLayer((28, 28, 1)),
-                tf.keras.layers.BatchNormalization(),
-                tf.keras.layers.Conv2D(32, (3, 3), padding="same"),
-                tf.keras.layers.ReLU(),
-                tf.keras.layers.MaxPooling2D((2, 2)),
-                tf.keras.layers.Conv2D(64, (3, 3), padding="same"),
-                tf.keras.layers.ReLU(),
-                tf.keras.layers.MaxPooling2D((2, 2)),
-                tf.keras.layers.Flatten(),
-                tf.keras.layers.Dense(100, activation="relu"),
-                tf.keras.layers.Dense(10, activation="softmax"),
+                keras.layers.InputLayer((28, 28, 1)),
+                keras.layers.BatchNormalization(),
+                keras.layers.Conv2D(32, (3, 3), padding="same"),
+                keras.layers.ReLU(),
+                keras.layers.MaxPooling2D((2, 2)),
+                keras.layers.Conv2D(64, (3, 3), padding="same"),
+                keras.layers.ReLU(),
+                keras.layers.MaxPooling2D((2, 2)),
+                keras.layers.Flatten(),
+                keras.layers.Dense(100, activation="relu"),
+                keras.layers.Dense(10, activation="softmax"),
             ]
         )
         self.do_test(nn)
@@ -87,12 +86,12 @@ def do_relu_tests(self, **kwargs):
         self.x_train = np.reshape(self.x_train, (-1, 28 * 28))
         self.x_test = np.reshape(self.x_test, (-1, 28 * 28))
 
-        nn = tf.keras.models.Sequential(
+        nn = keras.models.Sequential(
             [
-                tf.keras.layers.InputLayer((28 * 28,)),
-                tf.keras.layers.Dense(50),
-                tf.keras.layers.ReLU(**kwargs),
-                tf.keras.layers.Dense(10),
+                keras.layers.InputLayer((28 * 28,)),
+                keras.layers.Dense(50),
+                keras.layers.ReLU(**kwargs),
+                keras.layers.Dense(10),
             ]
         )
         self.do_test(nn)

tests/test_keras/test_keras_formulations.py

@@ -1,6 +1,6 @@
 import os
 
-import tensorflow as tf
+import keras
 from joblib import load
 
 from ..fixed_formulation import FixedRegressionModel
@@ -18,7 +18,7 @@ def test_diabetes_keras(self):
         X = load(os.path.join(self.basedir, "examples_diabetes.joblib"))
 
         filename = os.path.join(self.basedir, "diabetes.keras")
-        regressor = tf.keras.models.load_model(filename)
+        regressor = keras.saving.load_model(filename)
         onecase = {"predictor": regressor, "nonconvex": 0}
         self.do_one_case(onecase, X, 5, "all")
         self.do_one_case(onecase, X, 6, "pairs")
@@ -30,7 +30,7 @@ def test_diabetes_keras_alt(self):
             os.path.dirname(__file__), "..", "predictors", "diabetes_v2.keras"
         )
         print(filename)
-        regressor = tf.keras.models.load_model(filename)
+        regressor = keras.saving.load_model(filename)
         onecase = {"predictor": regressor, "nonconvex": 0}
         self.do_one_case(onecase, X, 5, "all")
         self.do_one_case(onecase, X, 6, "pairs")