TensorFlow Conv2D with kernel as input

Question

Is there a layer or function in TensorFlow that acts exactly like tf.keras.layers.Conv2D except the kernel/filter is an input to the layer? I am aware that