Closed guilherme-pombo closed 2 months ago
In what way do you mean it's not close at all?
The code will be released shortly, I'll let you know when it's uploaded! Best Wishes, Tom
Tom Joy DPhil Candidate in Machine Learning
thomas.joy@eng.ox.ac.uk (https://link.getmailspring.com/link/2BA3C71F-0A59-4B1A-8BA0-852C36552C46@getmailspring.com/0?redirect=%E2%80%9Cmailto%3Athomas.joy%40eng.ox.ac.uk%E2%80%9D&recipient=cmVwbHkrQUZVUTNQTkNWUUVONkhLVktYQkhCTE41VDJOQkxFVkJOSEhDV1pTWDRFQHJlcGx5LmdpdGh1Yi5jb20%3D) https://thwjoy.github.io (https://link.getmailspring.com/link/2BA3C71F-0A59-4B1A-8BA0-852C36552C46@getmailspring.com/1?redirect=https%3A%2F%2Fthwjoy.github.io&recipient=cmVwbHkrQUZVUTNQTkNWUUVONkhLVktYQkhCTE41VDJOQkxFVkJOSEhDV1pTWDRFQHJlcGx5LmdpdGh1Yi5jb20%3D)
On Oct 23 2020, at 12:05 pm, guilherme-pombo notifications@github.com wrote:
Hello, I've tried to reproduce the training algorithm using the details from your paper and I haven't been able to get close to the results you present on it. It would be very helpful if you could post the code for it or at least some pseudo-functioning code for it? Fashion-MNIST seems to work ok, but for CelebA it's not close at all. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub (https://link.getmailspring.com/link/2BA3C71F-0A59-4B1A-8BA0-852C36552C46@getmailspring.com/2?redirect=https%3A%2F%2Fgithub.com%2Fthwjoy%2Frevae-demo%2Fissues%2F2&recipient=cmVwbHkrQUZVUTNQTkNWUUVONkhLVktYQkhCTE41VDJOQkxFVkJOSEhDV1pTWDRFQHJlcGx5LmdpdGh1Yi5jb20%3D), or unsubscribe (https://link.getmailspring.com/link/2BA3C71F-0A59-4B1A-8BA0-852C36552C46@getmailspring.com/3?redirect=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAFUQ3PMHXHKXTZ57EIOEM2DSMFPRLANCNFSM4S4NDPZQ&recipient=cmVwbHkrQUZVUTNQTkNWUUVONkhLVktYQkhCTE41VDJOQkxFVkJOSEhDV1pTWDRFQHJlcGx5LmdpdGh1Yi5jb20%3D).
I've got it a bit closer now, but basically the interventions I do on CelebA, don't produce images as coherent as the ones presented on your paper. Look forward to seeing the nuances of your training code
Hi Tom,
Do you have already an idea when you'll be able to share the training code for reproducability of your paper?
Hi Adrian,
Yes it should be up next week sometime, the reason for the late upload is we were waiting for it to be accepted into ICLR, see new paper: https://openreview.net/forum?id=wQRlSUZ5V7B.
Best Wishes, Tom
Tom Joy DPhil Candidate in Machine Learning
thomas.joy@eng.ox.ac.uk (mailto:thomas.joy@eng.ox.ac.uk) http://thwjoy@github.io
On Mar 8 2021, at 1:26 pm, adriaan-lambrechts notifications@github.com wrote:
Hi Tom, Do you have already an idea when you'll be able to share the training code for reproducability of your paper? — You are receiving this because you commented. Reply to this email directly, view it on GitHub (https://github.com/thwjoy/revae-demo/issues/2#issuecomment-792754662), or unsubscribe (https://github.com/notifications/unsubscribe-auth/AFUQ3PJBGHDPWQX26N6RCLLTCTF7PANCNFSM4S4NDPZQ).
Hi All,
Apologies for the long delay, the training code for our ICLR work is here: https://github.com/thwjoy/ccvae.
Enjoy!
Hello, I've tried to reproduce the training algorithm using the details from your paper and I haven't been able to get close to the results you present on it. It would be very helpful if you could post the code for it or at least some pseudo-functioning code for it? Fashion-MNIST seems to work ok, but for CelebA it's not close at all.