Keras incompatible shapes NN - tensorflow

Question

So I have this neural network and I am feeding examples "X" and labels "Y" whose shapes are:
X.shape = (10,10,2)
Y.shape = (10,10,2)
The code for the model looks like:
import tensorflow as tf
from convert import process
import numpy as np
X, Y, rate = process('songs/song1.wav')
X = np.array(X[:10])
Y = np.array(Y[:10])
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(128))
model.add(tf.keras.layers.Dense(128))
model.add(tf.keras.layers.Dense(20))
model.compile(optimizer='adam', loss='categorical_crossentropy')
model.fit(X, Y, epochs=2)
Now for some reason once I run this i get the error:
ValueError: Shapes (None, 10, 2) and (None, 20) are incompatible
I am confused because I fed it data where each example of both "X" and "Y" have shapes (10, 2). So why is it saying that I passed it (None, 10, 2) and (None, 20)

Answer 1

Your last layer uses linear activation whereas you choose categorical_crossentropy loss. Set either
model.add(tf.keras.layers.Dense(20, activations='softmax'))
....loss='categorical_crossentropy')
or,
model.add(tf.keras.layers.Dense(20))
....loss='mse')
Also check your data shape, especially the label (y).