Keras : AttributeError: 'int' object has no attribute 'ndim' when using model.fit

Question

when triying to fit the model i get this error

i'm using Keras and every time i try to fit my model

padded_model.fit(train_X, train_y, epochs=50, verbose=1)

i get this error :

'int' object has no attribute 'ndim'

Solution

If train_x and train_y are normal Python lists, they don't have the attribute .ndim. Only Numpy arrays have this attribute representing the number of dimensions.

(https://docs.scipy.org/doc/numpy-1.12.0/reference/generated/numpy.ndarray.ndim.html)

Comments

[umutto]

What is train_x and train_y? They need to be numpy arrays or list of numpy arrays.

[collin]

@umutto i've got train_c and train y using train_test_split : 'train_X, test_X, train_y, 'test_y = train_test_split(padded_docs, y, train_size=0.7, test_size=0.3, random_state=123)'

[Dr. Snoopy]

Please include the types of train_X and train_y using the type() function in python.

[collin]

So the Problem is that the function fit cannot treate lists ?

[Joe]

Basically yes, but it is also not recommended to use lists. Numpy is a special package to speed up numerical calculations. Using pure Python lists is much slower and not recommended for this purpose. But it is easy to turn a list into a Numpy array if the module is installed. Just pass the list to the array constructor: x = np.array([1.0, 2.0, 3.0])