I have a multiclass problems. Three classes with labels 0 1 and 2.
If I create a model netowrk and I run this at the end
model.add(Activation('softmax'))
model.compile(optimizer = 'adam',loss='sparse_categorical_crossentropy',
metrics=['sparse_categorical_accuracy',tf.keras.metrics.AUC(name = 'auc')])
I always receive immediately the error
ValueError: Shapes (None, 3) and (None, 1) are incompatible
If I remove
tf.keras.metrics.AUC(name = 'auc')
from the metrics everything works.
I read the documentation about tf.keras.metrics.AUC
and it should work in the case of multi-class, but to me it does not.