-
Notifications
You must be signed in to change notification settings - Fork 21
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Breaking the Equivariant when using the haiku.module class #23
Comments
Yeah I believe this is because in Haiku, the init gets called on every forward pass of the model. You can check by calling the above If you want to use it inside another Haiku module, my advice would be either
|
Thanks for your early reply! I will try it soon. |
Hello, it's a great project!
I tried to use the
EMLP
with dm-haiku, and I write two version of codes in different ways. The first is directly using theemlp.nn.haiku
, and the second is using thehaiku.module
class. But I found that the second version will break the equivariance of neural network. In my view, the two versions have no difference in the architecture of neural network. Could you tell me if something wrong with my way of usingEMLP
?Attached is the codes. The first is:
and the second is:
The text was updated successfully, but these errors were encountered: