HareeshBahuleyan/tf-var-attention

running errors

houze-liu opened this issue · 1 comments

my tensorflow version is 1.12.0 and got somes errors when running your code

if not rnn_cell_impl._like_rnncell(cell): # pylint: disable=protected-access

if not rnn_cell_impl._like_rnncell(cell): # pylint: disable=protected-access

rnn_cell_impl._like_rnncell() seems no longer exists. Instead, we got:
https://github.com/tensorflow/tensorflow/blob/d63e3ea9a26fc049c654a966d0ebc56bc2747729/tensorflow/python/ops/rnn_cell_impl.py#L65
(assert_like_rnncell() won't allow passing for DropoutWrapper type, which is the type of cell above when running your code)

Another error:

layer_output_shape = self._output_layer._compute_output_shape( # pylint: disable=protected-access

compute_output_shape works instead of _compute_output_shape; didn't check source code.
I'm not very sure if my problem relating to 'pylint', but as far as I know # pylint: disable=protected-access is just used to disable warning messages.

Hi @houze-liu
The code in this repository was originally written in tensorflow==1.3.0 (please check README for other package version requirements)
I haven't tested it for the latest version of tensorflow. There could be some compatibility issues such as the ones you have encountered.

Could you please test it on tensorflow 1.3.0 instead?

Hareesh