CODEWITHSUNDEEP
pythonarraysnumpykeras
user963241
user963241

Reputation: 7048

What are argument types for fit() function in keras?

According to the Keras documentation:

fit(x=None, y=None, batch_size=None, epochs=1, ...)

The arguments for fit function are numpy arrays:

Arguments

x: Numpy array of training data (if the model has a single input), or list of Numpy arrays (if the model has multiple inputs). If input layers in the model are named, you can also pass a dictionary mapping input names to Numpy arrays. x can be None (default) if feeding from framework-native tensors (e.g. TensorFlow data tensors).

y: Numpy array of target (label) data (if the model has a single output), or list of Numpy arrays (if the model has multiple outputs). If output layers in the model are named, you can also pass a dictionary mapping output names to Numpy arrays. y can be None (default) if feeding from framework-native tensors (e.g. TensorFlow data tensors).

But I can pass x and y as dataframe and series.

print(type(x_train))
print(type(y_train))

Output:

<class 'pandas.core.frame.DataFrame'>
<class 'pandas.core.series.Series'>

model.fit(x_train, y_train, ..)

This works fine. Why is that?

Upvotes: 1

Views: 1682

Answers (1)

today
today

Reputation: 33460

Actually, this case has been covered in Keras source code by standardizing the input data in these lines:

    # Case 2: Symbolic tensors or Numpy array-like.
    x, y, sample_weights = self._standardize_user_data(
        x, y,
        sample_weight=sample_weight,
        class_weight=class_weight,
        batch_size=batch_size)

The _standardize_user_data method calls standardize_input_data function which converts DataFrame to Numpy array using using .values attribute of DataFrame in this line:

    data = data.values if data.__class__.__name__ == 'DataFrame' else data

As for Series, this is done a bit more implicitly in standardize_single_array function in this line:

    elif x.ndim == 1:
        x = np.expand_dims(x, 1)

Upvotes: 2

Related Questions