alandongmu/timegan_pytorchrebuild

Use pytorch to rebulid TimeGAN.

timegan_pytorchrebuild's Issues

sine dataset visualization problem

Hello author, when I ran your code to verify the sine data set, there was a problem with the visualization, that is, the generated data and the original data were not distributed in the same distribution. I am not very good at coding and would like help finding the source of the problem.
The resulting visualization example is as follows:


The original paper is visualized as follows：

Joint Training

Hi, thanks for your great work!
I wonder if you miss "self.E_loss.backward()" after this line:
https://github.com/AlanDongMu/TimeGAN_PytorchRebuild/blob/master/timegan.py#L88C14-L88C14

How can I change the num_workers in data_loading.py

How can I get the new generated data (with .csv format)?

Sorry for my naive question, I am a beginner of deeplearning.
When I run your code successfully, there is only PCA and t-sne test image in folder ''output'' .
So how can I get the new generated data with .csv format?
Thank u very much!

Error running main on defaults

When running main with the defaults, I receive this error regarding line 47 of run.py and line 89 in train_embedder():

RuntimeError: Trying to backward through the graph a second time (or directly access saved tensors after they have already been freed). Saved intermediate values of the graph are freed when you call .backward() or autograd.grad(). Specify retain_graph=True if you need to backward through the graph a second time or if you need to access saved tensors after calling backward.

I have tried setting retain_graph=True but the error does not go away.