search

Enny1991 / PLSTM

137 stars 31 forks source link

how to remember states with multi layer PLSTM #2

Closed vivanov879 closed 7 years ago

vivanov879 commented 7 years ago

Hi, How to remember states of lstm between runs?

Enny1991 commented 7 years ago

Hi, Can you clarify what you mean with "remembering states"? Do you mean that you would like to have access to the states of all the layers?

indiejoseph commented 7 years ago

it's seems no way to pass the initial state to multiPLSTM, how can i resume the previous rnn state? Or can i use MultiRNNCell with PhasedLSTMCell directly?

vivanov879 commented 7 years ago

made a pull request

vivanov879 commented 7 years ago

https://github.com/Enny1991/PLSTM/pull/3

Enny1991 commented 7 years ago

@indiejoseph I'm not sure is gonna be possible to use MultiRNNCell because we need to add the time input between layers. But now you can use multiPLSTM to resume the previous state.

vivanov879 commented 7 years ago

thanks a lot -- all's working -- now running your code : /usr/bin/python3.5 /home/vivanov/PycharmProjects/PLSTM/simplePhasedLSTM.py Compiling RNN... DONE! Compiling cost functions... DONE! /usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/gradients_impl.py:91: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory. "Converting sparse IndexedSlices to a dense Tensor of unknown shape. " Calculating gradients... DONE! Initializing variables... DONE! +-----------+----------+------------+ | Epoch=0 | Cost | Accuracy | +===========+==========+============+ | Train | 0.482815 | 0.785156 | +-----------+----------+------------+ | Test | 0.245212 | 0.9375 | +-----------+----------+------------+