Open tuanshanyou opened 7 years ago
@tuanshanyou - Please tell me the problem/error you are facing exactly? Though I was thinking to write down the explaination of the code.
Thank you for your reply. its requirements include python 3.5,but it has some grammar which is delete or modify in 3.5 such as "xrange and reload(sys)" so i use python2.7 to try it .
then the question is " Input 'split_dim' of 'Split' Op has type float32 that does not match expected type of int32." in model.py"encoder_inputs_split = tf.split(0, self.args.num_steps, XR)" I exchanged 1 and 2 position to solve it .(it is tensorflow1.0's grammar)
then the question i meet is "cannot import name rnn_cell". it shold be the version problem of tensorflow. in tensorflow1.0(which i use now ) ,it may be "tf.contrib.rnn.core_rnn_cell",not " tensorflow.models.rnn.rnn_cell". i change it into "from tensorflow.contrib import rnn ,rnn.BasicLSTMCell /n rnn.LSTMCell(...)...".and change "tf.nn.rnn(cell, encoder_inputs_split, dtype=tf.float32)" into "tf.nn.dynamic_rnn(cell, encoder_inputs_split, dtype=tf.float32)".the former works but th later lacks a parameter. then I decided to put it aside for the time being..
Thanks @tuanshanyou for pointing it out. The code was written with tensorflow v0.12. I will update the code to tensorflow v1.0.
the structure is complicated。 it makes a lot of unexpected mistakes...