Closed shane-huang closed 7 years ago
Hi there and thanks for pointing this out. I think the whole RNN-api has been restructured for the better and the TF-community has developed more good habits w.r.t how tf-code is written, something the ipynb at current doesn't reflect so the best is to just delete it.
Check out the branch wtte-rnn-0.0.2 and in particular the new examples.
The data-pipeline is something of a WIP but I think the discretely padded pipe is working quite good, I just haven't gotten a scalable RNN with varying-length sequences down yet without using batch_size=1.
Would dearly welcome PRs and comments and don't be shy to (especially) comment on style/naming etc and structure of the repo.
A few tasks on my mind has been:
Passes with tensorflow 1.0.1
I'm running tensorflow 1.0.1 and tried the ipython notebook in the example folder. But it break at several points. It seems code is broken because of incompatibility of tensorflow versions. Which version of tensorflow have you used to developed your code? And is it possible to migrate this to Tensorflow 1.0?
Thanks for the great work by the way. It's useful.