Why does keras complain about the loss function?

Question

Python

there is a keras model that is saved via model.save() This model uses a self-written loss function:

def contrastive_loss(y_true, y_pred):
        return K.mean(y_true * K.square(y_pred) + (1 - y_true) * K.square(K.maximum(1 - y_pred, 0)))
model.compile(loss=contrastive_loss, optimizer=rms)

There are no problems when training the model, but when loading the model:

from tensorflow.keras.models import load_model
model = load_model("model.h5")

An error is flying: ValueError: Unknown loss function:contrastive_loss
Please tell me how to fix this