0.7.1 version requires another parameter for initialising the LSTM cell (https://github.com/tensorflow/tensorflow/blob/r0.7/tensorflow/python/ops/rnn_cell.py#L264) which is the input_size for the cell. In versions 0.8 and beyond, they assume the input_size to be equal to num_hidden if not explicitly mentioned.
You could use this line of code with 0.7.1,
cell = rnn_cell.LSTMCell(num_hidden,num_hidden)
or upgrade to 0.8 or later.