search

martin-gorner / tensorflow-rnn-shakespeare

Code from the "Tensorflow and deep learning - without a PhD, Part 2" session on Recurrent Neural Networks.
Apache License 2.0
534 stars 249 forks source link

Continue training checkpoint #31

Closed kertase closed 6 years ago

kertase commented 6 years ago

I am not quite sure if this is possible, I might just not have understood the doc well.

So is it possible to continue training an previous checkpoint?

martin-gorner commented 6 years ago

Not with the code as it is now. If you look at the rnn_train.py file, you will see the line that saves a checkpoint but there is no line that restores it. Automatic checkpoint save/restore is offered by the Estimator interfece in Tensorflow but that interface does not yet support stateful RNNs like the one we are training here, where state must be passed around at each training step.