Cannot reproduce a layout about blobgan HOT 4 CLOSED

Yes, the shape of covs is actually not N x M x 2 x 2 unless covs_raw is False, sorry about that. If you run the code you'll see it's an N x M x 4 packed tensor of all the blob parameters which is then processed into the final covariance matrix. Apologies for the misleading code segment!

from blobgan.

Hi @dave-epstein,

Thanks for the explanation. I run the following code:

layout, img = model.gen(layout=layout, ret_layout=True)

It turns out that the bug appeared (that NM22 is outputted) when the layout is provided. As a result, a simple reshape from NM22 to NM4 would work, right?

from blobgan.

That should not be necessary. I've run that code snippet that you sent without any problem. See this line which takes care of the correct parameter settings automatically: https://github.com/dave-epstein/blobgan/blob/main/src/models/blobgan.py#L183

Do you have an old version of the code or something like that?

from blobgan.

It worked after updating the code to the latest version. Thanks.

from blobgan.