Dense layer only supports RaggedTensors when the innermost dimension is non-ragged. Received: inputs.shape=(None, None, None)

Question

I have a tf.data.dataset with 2 and 3 dimensional ragged tensors:

[
 [1,2,3,4],
 [5,6,7,8],
 ...
] 
and [
     [[1,2,3],
      [4,5,6]
     ],
     [[7,8,9],
      [10,11,12],
      ...
     ]
    ]

I receive the error in the title when passing the inputs to a Dense layer. The Ragged Tensor guide https://www.tensorflow.org/guide/ragged_tensor has ragged tensors in the same shape as my 2D RaggedTensors and they use Dense layers just fine, despite the inner-most dimension being ragged, so I am not sure why mine is not working.

Any advice?