Apparently this is working as intended - in general models should be sourced from the repo, not the pip package.

So are our instructions incorrect somewhere? We should update them if that's the case.

People get confused because build_pip_package imports models/rnn/{rnn,rnn_cell,seq2seq}. If/when we move these into python.ops.nn (e.g. something like nn.rnn, nn.rnn_cell, nn.seq2seq); that should fix some of the confusion.

Agreed. Until we are ready to support those interfaces indefinitely, is there something we should do to avoid that confusion? Should we remove models/rnn from the pip package?

No, they are the proper rnn/lstm implementation and should be made
available with the rest of the ops.
On Nov 10, 2015 1:08 PM, "Vijay Vasudevan" notifications@github.com wrote:

Agreed. Until we are ready to support those interfaces indefinitely, is
there something we should do to avoid that confusion? Should we remove
models/rnn from the pip package?

—
Reply to this email directly or view it on GitHub
#69 (comment)
.

is this the reason why the rnn/lstm is not in the documentation under nn (in the api doc)?

Just for the time being, what's the best resources to learn how to use them? I've just been reading the rnn and seq-to-seq tutorial, as there is no documentations for these modules.