FactorVAE

FactorVAE copied to clipboard

Please give versions of torchvision and tqdm and visdom

swainsubrat

Dependency configuration and RuntimeError

3

I don't think the README lists all of the required dependencies; in particular, `torchvision` seems to be a required import to run the training scripts. However, `torchvision` is only compatible...

jpgard

Hello! The error is as follows, please advise. Checking for scripts. It's Alive! INFO:root:Application Started You can navigate to http://localhost:8097 INFO:tornado.access:200 POST /env/main (127.0.0.1) 0.50ms INFO:tornado.access:101 GET /vis_socket (127.0.0.1) 0.29ms...

DXSHAN-NEU

Why is 64x64 the only size allowed and would I be able to change this?

gene-chou

Disentanglement metric

2

Issue #9 was closed, however the mentioned loss metric, to my understanding, didn't get pushed to the repo. Would it be possible for you to do this?

FrankBrongers

Inplace problem when calling D_tc_loss.backward()

1

the error message is: `one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [4096, 6]], which is output 0 of TBackward, is at...

anewusername77

D_tc_loss line 119 in solver.py

1

I am not sure about the following code, can someone help to explain it? D_tc_loss = 0.5*(F.cross_entropy(D_z, zeros) + F.cross_entropy(D_z_pperm, ones))

goodman1204

I'm running on the resized (1,64,64) chairs, and it takes 4.5 hrs to get to 5% of training. I ran with the given values and my cuda() is being used...

karamarieliu

The vae tc loss

1

HI, Thanks your code, its very useful. I notice that the tc_vae_loss in solver.py. It could be a negative number, and this is unusual for loss functions.

cuicanyu