1

I run a hybrid model in keras (external library). It keeps on running for more than 24 hours. However, the program is not utilizing the GPU. How can I transform my program to run on tensorFlow environment using tensorFlow.keras to utilize my GPU? following is the code:

model = keras.Sequential(
    [
        keras.layers.InputLayer(input_shape=(231, 231, 3)),
        layers.Conv2D(128, 3, padding="same", activation="relu"),
        layers.Conv2D(128, 3, padding="same", activation="relu", kernel_initializer = 'glorot_uniform'),
    ]
)

extraction_model = Model(model.input, model.layers[1].output)
new_X = extraction_model.predict(X)

x_train = new_X.reshape(-1, new_X.shape[3]) 
y_train = Y.reshape(-1)

RF_clf = RandomForestClassifier(random_state=42, oob_score=True)
SV_clf = SVC(random_state=42, probability=True)
LR_clf = LogisticRegression(random_state=42,)

estimators = [('RF', RF_clf),  ('SV', SV_clf), ] 

clf = StackingClassifier(estimators=estimators, final_estimator=LogisticRegression())
clf.fit(x_train, y_train)
print("Stacking model score: %.3f" % clf.score(x_test, y_test))
T. Elhassan
  • 23
  • 5
  • Does this answer your question? [How do I use TensorFlow GPU?](https://stackoverflow.com/questions/51306862/how-do-i-use-tensorflow-gpu) – Proko Aug 12 '21 at 22:46

1 Answers1

0

the Keras was already built under the hood of Tensorflow. if you train your model ad use the training functions, it will use GPU if you have a GPU available with all Cuda and dependency installed.

if you are still worried about the GPU issue, better use the google colab for running the training which can help you in many ways.

Tamil Selvan
  • 1,600
  • 1
  • 9
  • 25