You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@IshmaelBelghazi@vdumoulin I am curious what can be the reason not to provide pretrained models? The code works perfectly well and I really do not see a single reason to force people who want to compare to your work (and cite) spend time on installing an unusual framework and waiting for a day while the code runs.
@DmitryUlyanov Rest assured, we are not intentionally withholding trained models. This is simply an oversight on our part.
I 100% agree that providing trained models would be the more user-friendly thing to do; unfortunately the only trained model files we have lying around internally are from an older, pre-open-sourced version of the code and are not compatible with the current version of the code, which means I would also have to re-train the models just like you.
With the NIPS deadline coming I won't have time to do so until at least next week, but in the meantime if you happen to have trained models using the open-source version of the code and would be willing to contribute back we would make sure to credit you for the effort. :)
Hi, I am so interested with your work, could you upload your pre-trained models (i.e., ImageNet) online ? Thanks!
The text was updated successfully, but these errors were encountered: