Keras/Tensorflow model seems to run the 'call' method twice on fit - why?

I added a print to a call function in Keras model, and suspect that the call method is called twice on model.fit - what is the reason for that?

class my_model(keras.Model):
    def __init__(self,  units = 3, **kargs):
        super().__init__(**kargs)
        self.hidden = keras.layers.Dense(1)
    
    def call(self, inputs):
        outputs = self.hidden(inputs)
        print ('my_model call function')
        return outputs

model = my_model()
model.compile(loss = "mse")

X = np.random.standard_normal([10,3])
y = np.random.standard_normal([10,1])

history = model.fit(X, y, epochs = 1000, verbose = 0)

Output:

my_model call function
my_model call function