Bugfix: now works correctly even if the input Tensor(s) do not require grad.
This works by adding a fake input to the forward()
of torch.autograd.Function
.
Also adds an automatic detection of prologue (first layers with no trainable parameters)
La mise à jour de gitlab est terminée. Nous sommes désormais en version 16.11.1
Merci de consulter la release note:
https://about.gitlab.com/releases/2024/04/18/gitlab-16-11-released/
This works by adding a fake input to the forward()
of torch.autograd.Function
.
Also adds an automatic detection of prologue (first layers with no trainable parameters)