Keras模型fit_generator奇数错误

Question

I have the following code:

    datagen = ImageDataGenerator(
        rescale=1./255,
        target_size=(128, 128),
        shear_range=0.2,
        zoom_range=0.2,
        horizontal_flip=True)

    test_datagen = ImageDataGenerator(
        rescale=1./255,
        target_size=(128, 128)
    )

    datagen.fit(X_train)

    model.fit_generator(
        datagen.flow(X_train, Y_train),
        samples_per_epoch=len(X_train),
        epochs=30,
        verbose=1,
        validation_data=(X_valid, Y_valid))

Which throws this unusual error

  Traceback (most recent call last):
      File "cnn.py", line 258, in <module>
          models = run_cross_validation_create_models(num_folds)
      File "cnn.py", line 205, in run_cross_validation_create_models
          validation_data=(X_valid, Y_valid))
      TypeError: fit_generator() takes at least 4 arguments (5 given)

Can somebody explain what is going wrong here, I am loading in a set of 3700 images.

Answer 1

It might come from the new API (Keras 2.0 released yesterday), the fit_generator() now takes steps_per_epoch argument instead of samples_per_epoch .

The steps_per_epoch is typically samples_per_epoch / batch_size .

You can find this info in the documentation .

Does it help?