Hi @yang-song

I use the following code to test the BasicBlock, but I found the determinant is always 0. Is this normal? I am looking forward to your reply.

basic_model = BasicBlock(
        input_dim=3,
        latent_dim=32,
        kernel1=3,
        kernel2=3,
        kernel3=3,
        type='A',
        config=config,
        shape=[3, 32, 32],
        init_zero=False,
    )
x = torch.randn(1, 3, 32, 32)
log_det = torch.zeros(1)
out, log_det = basic_model([x, log_det])
print(log_det)

Out: tensor([0.], grad_fn=<AddBackward0>)

Hey, thank you for sharing the code. It's great!
I had one query. I have trained a Mintnet model for classification on CIFAR10. Can I use this same model for generating new samples i.e as a generative model (with some code modifications, of course)? If yes, can you suggest what part of the code needs to be changed?
As far as I understood, currently, density estimation and classification experiments are run independently, but I want to make changes to achieve the above use-case.
Thanks in advance!

Hi @yang-song,
Thanks for releasing the code, I really enjoy your paper and would like to use the invertibility in our work.
However, I cannot find the inverse functions in the MintNet class or Mint Block in your code.
Also, how can we reproduce Fig.4a (verification of invertibility)?
Thanks

Hello! I have really enjoyed reading your paper and want to do some experiments with the invertible network on Cifar-10. However, I think there are some small mistakes with the CIFAR-10 config file, namely that it refers to MNIST and has the wrong image size (28 instead of 32). After rectifying this, I get another error, namely

File "/home/kyle/mintnet/models/cnn_classification.py", line 122, in forward
    center2 = center2.permute(0, 2, 1, 3, 4, 5).contiguous().view_as(self.weight2)
RuntimeError: shape '[16, 16, 1, 1]' is invalid for input of size 2304

I'm guessing some other parameter in this config file is not correct, but I couldn't determine which one from the paper. Would you be able to help me figure it out?

Additionally, would you happen to have checkpoints for the trained CIFAR10 model lying around? At least for now, I don't actually need to train a model myself, I just want to use a trained one.

Thanks in advance!