jeremycchsu / vae-npvc Goto Github PK
View Code? Open in Web Editor NEWRe-implementation the code used in Voice Conversion from Non-parallel Corpora Using Variational Auto-encoder
License: Other
Re-implementation the code used in Voice Conversion from Non-parallel Corpora Using Variational Auto-encoder
License: Other
https://github.com/JeremyCCHsu/vae-npvc/blob/master/model/vae.py#L128
according to paper, loss['G'] = logPx - D_KL??
As you said my tensorflow is working on cpu , so I changed format='NHWC' still getting same error:Conv2DCustomBackpropInputOp only supports NHWC.
so kindly tell me in which file I need to change.
The results on VCTK dataset are good.However when i use it for my own dataset the reconstructions are very noisy and inaudible.Any suggestions?
Hi @JeremyCCHsu
Could you guide me on this convertion? I have no idea which ops are nchw, and which are nhwc.
Thank you.
我想使用中文数据进行训练,需要改什么吗?还是说直接把数据集改了就可以,其余的代码都不用改
Hi,
I am trying to do non-parallel voice conversion, with VAE option I am able to do voice conversion. But when I changed model to VAWGAN, it is giving below error.
python main.py --model VAWGAN --trainer VAWGANTrainer --architecture architecture-vawgan-vcc2016.json
FutureWarning: Conversion of the second argument of issubdtype from float
to np.floating
is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type
.
from ._conv import register_converters as _register_converters
Traceback (most recent call last):
File "main.py", line 40, in
MODEL = getattr(module, args.model)
AttributeError: module 'model.vae' has no attribute 'VAWGAN'
root@vultr:~/vae-npvc# python3 main.py --model ConvVAE --trainer VAETrainer --architecture architecture-vae-vcc2016.json
Using default logdir: logdir/train/0114-1444-11-2019
Traceback (most recent call last):
File "main.py", line 78, in
main()
File "main.py", line 58, in main
xmax=np.fromfile('./etc/xmax.npf'),
FileNotFoundError: [Errno 2] No such file or directory: './etc/xmax.npf'
Compiler tips can not find the file, can you answer, beginners do not understand
i am curious anout https://github.com/JeremyCCHsu/vae-npvc/blob/master/model/vae.py#L124
logPx = tf.reduce_mean(
GaussianLogDensity(
slim.flatten(x),
slim.flatten(xh),
tf.zeros_like(slim.flatten(xh))),
if log_var is constant 0, leading to loss GaussianLogDensity is equivalent to MSE?
In “main.py”, training data are got by " image, label = read(***)" for once. But I found the function "read()" in analyzer.py can only return a batch-size samples and there is no circulation for getting other samples. Thus, only a batch-size samples are used to train the network. Do I understand your code correctly? If yes, is it the right way to read training data?
Hi Jeremy, I am running convert.py and get this error:
InvalidArgumentError: Assign requires shapes of both tensors to match. lhs shape= [7,1,1,16] rhs shape= [7,1,513,16]
[[Node: save/Assign_1 = Assign[T=DT_FLOAT, _class=["loc:@Encoder/Conv2d-0/Conv2d-0/kernel"], use_locking=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/cpu:0"](Encoder/Conv2d-0/Conv2d-0/kernel, save/RestoreV2_1)]]
It is because I use CPU, so I change all the NCHW to NHWC to support CPU version.
An error named "ValueError: Dimensions must be equal, but are 513 and 41553 for 'loss/GaussianLogDensity/sub' (op: 'Sub') with input shapes: [16,513], [16,41553]."was occured ,when I run main.py using ConvVAE,VAETrain.Besides,my environment is py36,tf1.2 on cpu.Could you help me fix the problem?Thank you.
here is the message I've got.
--2017-12-14 11:32:50-- http://datashare.is.ed.ac.uk/download/10283/2042/SUPERS EDED_-The_Voice_Conversion_Challenge_2016.zip
Resolving datashare.is.ed.ac.uk (datashare.is.ed.ac.uk)... 129.215.41.53
Connecting to datashare.is.ed.ac.uk (datashare.is.ed.ac.uk)|129.215.41.53|:80... connected.
HTTP request sent, awaiting response... 302 Found : Moved Temporarily
Location: https://datashare.is.ed.ac.uk/download/10283/2042/SUPERSEDED_-_The_Voi ce_Conversion_Challenge_2016.zip [following]
--2017-12-14 11:32:52-- https://datashare.is.ed.ac.uk/download/10283/2042/SUPER SEDED-_The_Voice_Conversion_Challenge_2016.zip
Connecting to datashare.is.ed.ac.uk (datashare.is.ed.ac.uk)|129.215.41.53|:443.. . connected.
HTTP request sent, awaiting response... 404 Not Found
2017-12-14 11:32:52 ERROR 404: Not Found.
unzip: cannot find or open SUPERSEDED_-The_Voice_Conversion_Challenge_2016.zip , SUPERSEDED-The_Voice_Conversion_Challenge_2016.zip.zip or SUPERSEDED-_The_V oice_Conversion_Challenge_2016.zip.ZIP.
unzip: cannot find or open vcc2016_training.zip, vcc2016_training.zip.zip or vc c2016_training.zip.ZIP.
mv: cannot stat ‘vcc2016_training’: No such file or directory
unzip: cannot find or open evaluation_all.zip, evaluation_all.zip.zip or evalua tion_all.zip.ZIP.
rm: cannot remove ‘evaluation_all.zip’: No such file or directory
rm: cannot remove ‘vcc2016_training.zip’: No such file or directory
tensorflow.python.framework.errors_impl.InvalidArgumentError: Conv2DCustomBackpropInputOp only supports NHWC.
[[{{node Update/gradients/loss/Encoder/Conv2d-4/Conv2d-4/Conv2D_grad/Conv2DBackpropInput}}]]
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "F:/work/vae-npvc-master/main.py", line 78, in
main()
File "F:/work/vae-npvc-master/main.py", line 74, in main
trainer.train(nIter=arch['training']['max_iter'], machine=machine)
File "F:\work\vae-npvc-master\trainer\vae.py", line 99, in train
sess.run(self.opt['g'])
File "F:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\client\session.py", line 950, in run
run_metadata_ptr)
File "F:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\client\session.py", line 1173, in _run
feed_dict_tensor, options, run_metadata)
File "F:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\client\session.py", line 1350, in _do_run
run_metadata)
as the title, where can i listen to the demo?
I see there's wgan model in earlier commit, but now cleaned, how may I reproduce your result?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.