The python implementation of google tensor flow (program) states:
The default non-peephole implementation is based on:
http://deeplearning.cs.cmu.edu/pdfs/Hochreiter97_lstm.pdf
S. Hochreiter and J. Schmidhuber. "Long Short-Term Memory". Neural Computation, 9(8):1735-1780, 1997.
My question is: what is the equivalent of variable "num_units" on this paper (http://deeplearning.cs.cmu.edu/pdfs/Hochreiter97_lstm.pdf)?
Does it refer to one block, stated on the paper? Or one memory cell cj? Or what? The program itself says: "The number of units in the LSTM cell"!!
That is not clear for me when I try to map the paper to the program.
Thanks