RNN tutorial gives NotImplementedError

Using the latest version (2.0.3) DyNet in Mac OS, both python3.6 and python2.7 give this error on cpu. Install today with 'pip install dynet'
http://dynet.readthedocs.io/en/latest/tutorials_notebooks/RNNs.html
---------------------------------------------------------------------------
NotImplementedError Traceback (most recent call last)
in ()
103
104 sentence = "a quick brown fox jumped over the lazy dog"
--> 105 train(srnn, params_srnn, sentence)

in train(rnn, params, sentence)
94 trainer = dy.SimpleSGDTrainer(pc)
95 for i in range(200):
---> 96 loss = do_one_sentence(rnn, params, sentence)
97 loss_value = loss.value()
98 loss.backward()

in do_one_sentence(rnn, params, sentence)
54 s = s.add_input(lookup[char])
55 #print(R.dim,s.output().dim)
---> 56 probs = dy.softmax(R*s.output() + bias)
57 loss.append( -dy.log(dy.pick(probs,next_char)) )
58 loss = dy.esum(loss)

_dynet.pyx in _dynet.Expression.__mul__()

NotImplementedError:

Start with the RNN tutorial notebook at tutorials_notebooks/RNNs.html and reproduce the failure on DyNet 2.0.3 with the shown CPU setup. Inspect the multiplication at line 56 of do_one_sentence, then update the tutorial so the example runs without NotImplementedError and verify the notebook completes its training loop.

Written by the indexing model from the issue text.