process_batch_input_for_RNN could be shortened?

[wenyong-h]

    batch_input_ = tf.transpose(batch_input, perm=[2, 0, 1])
    X = tf.transpose(batch_input_)

should be equivalent to

    X = tf.transpose(batch_input, perm=[1, 0, 2])

[wenyong-h]

By the way, when i run the code under tensorflow 0.9, the loss and accuracy stay the same after every iteration.
the result is like this:

Iteration: 119 Loss: 230.362 Train Accuracy: 9.42857 Test Accuracy: 12.6263

I also tried to run under tensorflow 0.10, the loss will change a little bit, but could almost be ignored.

[KnHuq]

@wenyong-h Your loss is not moving because you are not getting any gradient. Please check the terminal when you run the jupyter notebook. Is there any error?

[wenyong-h]

@KnHuq i'm running the python file, not the ipython notebook. There was no error.
I tried using AdamOptimizer instead of AdadeltaOptimizer, then everything is right.
this is the result of one run of the Vanilla RNN

Iteration: 119 Loss: 46.1185 Train Accuracy: 99.6429 Test Accuracy: 94.697

[KnHuq]

@wenyong-h You are right! One transpose will do the same thing!