tfb.ScaleMatvecTriL does not work as expected

Hello, I found out that the bijectors "ScaleMatvecX" are not working as expected, at least according to the documentation (https://www.tensorflow.org/probability/api_docs/python/tfp/bijectors/ScaleMatvecTriL). According to this, the bijectors "Compute Y = g(X; scale) = scale @ X".
However this is not the case:

scale_tril = [[1., 0.], [1., 1.]]
X = tf.random.uniform((2, 4))

scale_tril @ X  # works as expected
tfb.ScaleMatvecTriL(scale_tril, validate_args=True)(X) # ValueError: Dimensions 2 and 4 are not compatible

Instead, the correct result occurs with

tfb.Chain([
    tfb.Transpose(rightmost_transposed_ndims=2),
    tfb.ScaleMatvecTriL(scale_tril, validate_args=True),
    tfb.Transpose(rightmost_transposed_ndims=2),
])(X) # works as expected, equivalent to scale_tril @ X

I'm not really sure what is happening in the bijector.
Thank you!

Start with the ScaleMatvecTriL entry point and reproduce the reported ValueError using the scale_tril and X examples in the issue. Compare its documented scale @ X behavior with the working Transpose/ScaleMatvecTriL/Transpose chain, then update the implementation or documentation so the behavior and documentation agree.