tf.GradientTape().gradient() returns None

Question

i tried to design my own loss function for a tensorflow 2.0 model.

def loss(y_true, y_pred):
    true_col = [col for col in y_true.columns if col not in [alt_id_c