Not to create unnessary tensorflow variables when sampling

sherjilozair / char-rnn-tensorflow

Multi-layer Recurrent Neural Networks (LSTM, RNN) for character-level language models in Python using Tensorflow

MIT License

2.64k stars 960 forks source link

Not to create unnessary tensorflow variables when sampling #44

Closed MingStar closed 8 years ago

MingStar commented 8 years ago

The tensorflow variable self.lr is really for traning only, as well as the variable self.train_op.

hunkim commented 8 years ago

I think these changes are interesting, but IMHO, there are too many features in one PR.

I guess it would be better if you can split them and submit separate PRs for each feature such as temperature, saving best results, web server, dropouts, etc.? I am also not sure having a web-server implementation in char-rnn is a good idea.

For the dropouts, is this similar to #35?

MingStar commented 8 years ago

Sorry. was meant to request for one commit only.

I will create a separate branch and edit the pull request.

MingStar commented 8 years ago

Will create a new pull request for this.

hunkim commented 8 years ago

Thanks!