The wrong dimension in discriminator function

[ShanHaoYu]

I found it there is a bug in discriminator function on tensorflow_MNIST_GAN.py.

The Original code is below: w2 = tf.get_variable('D_w2', [h1.get_shape()[1], 256], initializer=w_init) b2 = tf.get_variable('D_b2', [1024], initializer=b_init)

it should be modified as: w2 = tf.get_variable('D_w2', [h1.get_shape()[1], 256], initializer=w_init) b2 = tf.get_variable('D_b2', [256], initializer=b_init)

Many Thanks.

[aiXander]

Had the same bug, thx for posting, didn't feel like debugging dimension mismatches today ;p

[aiXander]

One extra thing: for people that are running Linux on Windows 10 (using developer mode): To be able to execute the show_result function you can install something like Xming and then use export DISPLAY=:0 to patch the results through!

[VictorZhang2014]

Change the original code 1024 to 256 will be just fine.

[vpshah3]

Hey I changed the size to 256 as mentioned but I am getting an error like this: ValueError: Trying to share variable D/D_b2, but specified shape (256,) and found shape(1024,)

Any idea on this??!!

[simk10]

same @vpshah3