ikostrikov / tensorflow-vae-gan-draw Goto Github PK

No module named 'progressbar'？from progressbar import ETA, Bar, Percentage, ProgressBar
where is progressbar?

Shouldn't it be reduced_mean instead of reduced_sum when computing the vae and reconstruction loss?

It seems to me that you are using the tensorflow deconv2d which performs a transpose convolution (see conv2d_transpose at https://github.com/tensorflow/tensorflow/blob/4c3b060b9b063a247172c939e0e1314577b581e3/tensorflow/python/ops/nn_ops.py). However, the deconv that we want to perform is an unpooling followed by convolution. Could you clarify whether my understanding is correct? Thanks.

Also relevant: tensorflow/tensorflow#256

Traceback (most recent call last):
File "main.py", line 43, in
model = VAE(FLAGS.hidden_size, FLAGS.batch_size, FLAGS.learning_rate)
File "/root/TensorFlow-VAE-GAN-DRAW/vae.py", line 51, in init
self.sess.run(tf.global_variables_initializer())
AttributeError: 'module' object has no attribute 'global_variables_initializer'

Hi,

Great work!

Usually in GAN, the generator is run twice to make sure that discriminator loss does not go to zero. And I noticed that you are not doing this in your implementation!

Thanks!

Hello,
I got this error: AttributeError: module 'mnist' has no attribute 'train_images'
when I ran this code:
import mnist
import itertools
import numpy as np

def prepare_data(images, labels):
images = map(lambda a: a / 255.0, map(np.asarray, map(list, map(itertools.chain.from_iterable, images))))
return list(zip(images, labels))

train = prepare_data(mnist.train_images(), mnist.train_labels())
test = prepare_data(mnist.test_images(), mnist.test_labels())

Can you please help me?

Hi, when I run the command 'python main-draw.py --working_directory /tmp/gan', I got the following error:

Traceback (most recent call last):
File "main-draw.py", line 190, in
gru_cell(num_units=FLAGS.rnn_size, state=pt.UnboundVariable('state')))
File "/home/amax/anaconda2/lib/python2.7/site-packages/prettytensor/pretty_tensor_class.py", line 2011, in method
return func(input_layer, *args, **self.fill_kwargs(input_layer, kwargs))
File "/home/amax/anaconda2/lib/python2.7/site-packages/prettytensor/recurrent_networks.py", line 286, in gru_cell
state = state[0]
TypeError: 'UnboundVariable' object does not support indexing
The version of my tensorflow is 1.0
I use "pip install prettytensor" to install prettytensor 0.7.4
when I tried the example from the tutorial : https://github.com/google/prettytensor, it failed too
A = (pretty_tensor.template('x')
.lstm_cell(num_units=256, state=UnboundVariable('state'))

Thanks !

Hi,

good work @ikostrikov ! I just wonder where the loss function:
https://github.com/ikostrikov/TensorFlow-VAE-GAN-DRAW/blob/master/main-gan.py#L60-L61
comes from?
Because it does not look like a cross entropy loss

I hope you do not mind my way of asking here.

When I try to run the code with prettytensor 0.7.3 and tensorflow 0.12.head, I get the error with no attribute tf.complex_abs, I assume that is because prettytensor doesn't support tensorflow 0.12.
However, when I try to run it with tensorflow 0.11.0rc1, I get the error no attribute 'GLOBAL_VARIABLES'. It seems like prettytensor requires tensorflow 0.12+.
Therefore I am stuck, are there any suggestions?

I downloaded the package and upgraded TensorFlow to v0.9. However, the first test run did not work:

python main-vae.py -- working_directory /tmp/gan
Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.
Extracting MNIST/train-images-idx3-ubyte.gz
Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.
Extracting MNIST/train-labels-idx1-ubyte.gz
Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.
Extracting MNIST/t10k-images-idx3-ubyte.gz
Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.
Extracting MNIST/t10k-labels-idx1-ubyte.gz
Traceback (most recent call last):
  File "main-vae.py", line 143, in <module>
    pbar = ProgressBar(max_value = FLAGS.updates_per_epoch, widgets=widgets)
TypeError: __init__() got an unexpected keyword argument 'max_value'

Apparently, something was changed in the ProgressBar package recently. Quick look at the examples revealed that maxval arg should be used rather than max_value.
Original:
pbar = ProgressBar(max_value = FLAGS.updates_per_epoch, widgets=widgets)
Modified:
pbar = ProgressBar(maxval = FLAGS.updates_per_epoch, widgets=widgets)
With this fix the package seems to work.

ENV(pip list)

tensorflow (0.12.0rc0)
prettytensor (0.6.2)

When I run this command

python main-vae.py

I encounter following problems. So I want to know what the env should be, or what should I do to fix the bug? Thanks for help.

WARNING:tensorflow:From main-vae.py:134 in .: initialize_all_variables (from tensorflow.python.ops.variables) is deprecated and will be removed after 2017-03-02.
Instructions for updating:
Use tf.global_variables_initializer instead.
Traceback (most recent call last):
File "main-vae.py", line 148, in
_, loss_value = sess.run([train, loss], {input_tensor: x})
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 766, in run
run_metadata_ptr)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 964, in _run
feed_dict_string, options, run_metadata)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1014, in _do_run
target_list, options, run_metadata)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1034, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.FailedPreconditionError: Attempting to use uninitialized value model/conv2d/weights
[[Node: model/conv2d/weights/read = IdentityT=DT_FLOAT, _class=["loc:@model/conv2d/weights"], _device="/job:localhost/replica:0/task:0/cpu:0"]]

Caused by op u'model/conv2d/weights/read', defined at:
File "main-vae.py", line 120, in
output_tensor, mean, stddev = decoder(encoder(input_tensor))
File "main-vae.py", line 42, in encoder
conv2d(5, 32, stride=2).
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/prettytensor/pretty_tensor_class.py", line 1981, in method
result = func(non_seq_layer, *args, **kwargs)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/prettytensor/pretty_tensor_image_methods.py", line 159, in call
params = self.variable('weights', size, init, dt=dtype)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/prettytensor/pretty_tensor_class.py", line 1695, in variable
collections=variable_collections)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 1024, in get_variable
custom_getter=custom_getter)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 850, in get_variable
custom_getter=custom_getter)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 346, in get_variable
validate_shape=validate_shape)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 331, in _true_getter
caching_device=caching_device, validate_shape=validate_shape)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 677, in _get_single_variable
expected_shape=shape)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/ops/variables.py", line 224, in init
expected_shape=expected_shape)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/ops/variables.py", line 370, in _init_from_args
self._snapshot = array_ops.identity(self._variable, name="read")
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/ops/gen_array_ops.py", line 1424, in identity
result = _op_def_lib.apply_op("Identity", input=input, name=name)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 759, in apply_op
op_def=op_def)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2240, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/home/tangluming/niuge/ENV/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1128, in init
self._traceback = _extract_stack()

FailedPreconditionError (see above for traceback): Attempting to use uninitialized value model/conv2d/weights
[[Node: model/conv2d/weights/read = IdentityT=DT_FLOAT, _class=["loc:@model/conv2d/weights"], _device="/job:localhost/replica:0/task:0/cpu:0"]]

Hi,

Awesome code!

A minor issue. Following paper https://arxiv.org/pdf/1312.6114.pdf, equation 10, it seems line 62 of vae.py

tf.reduce_sum(0.5 * (tf.square(mean) + tf.square(stddev) - 2.0 * tf.log(stddev + epsilon) - 1.0))

should be

tf.reduce_sum(0.5 * (tf.square(mean) + tf.square(stddev) - tf.log(stddev + epsilon) - 1.0))

envy@ub1404:/os_pri/github/TensorFlow-VAE-GAN-DRAW$ python main-vae.py --working_directory .
Extracting ./MNIST/train-images-idx3-ubyte.gz
Extracting ./MNIST/train-labels-idx1-ubyte.gz
Extracting ./MNIST/t10k-images-idx3-ubyte.gz
Extracting ./MNIST/t10k-labels-idx1-ubyte.gz
Traceback (most recent call last):
File "main-vae.py", line 120, in
output_tensor, mean, stddev = decoder(encoder(input_tensor))
File "main-vae.py", line 41, in encoder
reshape([FLAGS.batch_size, 28, 28, 1]).
File "/home/envy/.local/lib/python2.7/site-packages/prettytensor/pretty_tensor_class.py", line 1816, in method
with _method_scope(input_layer, scope_name) as (scope, _):
File "/usr/lib/python2.7/contextlib.py", line 17, in enter
return self.gen.next()
File "/home/envy/.local/lib/python2.7/site-packages/prettytensor/pretty_tensor_class.py", line 1628, in _method_scope
scopes.var_and_name_scope((name, None)) as (scope, var_scope):
File "/usr/lib/python2.7/contextlib.py", line 17, in enter
return self.gen.next()
File "/home/envy/.local/lib/python2.7/site-packages/prettytensor/scopes.py", line 57, in var_and_name_scope
vs_key[0] = variable_scope._VariableScope(
AttributeError: 'module' object has no attribute '_VariableScope'
envy@ub1404:/os_pri/github/TensorFlow-VAE-GAN-DRAW$

I found that you used elu as the activation function for gan, but on the DCGAN paper they suggest us to use lrelu for the discriminator, relu for encoder and decoder. Also elu is pretty new. I wonder if there is some logic behind your choice.