-
Notifications
You must be signed in to change notification settings - Fork 3
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
Add MLPMixer test #154
base: main
Are you sure you want to change the base?
Add MLPMixer test #154
Conversation
# Alternatively, weights could be randomly initialized like this: | ||
# weights = self._model.init(jax.random.PRNGKey(42), ins, train=False) | ||
|
||
return [weights, ins] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
As far as I can see, MlpMixer.__call__
has only two params inputs, train
. Where did [weights, ins]
come from? Does apply
use them?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes. The way flax linen works is that modules have an apply
method that takes an argument called variables
(representing parameters, and possibly other state like batch statistics) alongside real model inputs, and then binds the state to where it needs to go and forwards inputs to __call__
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah, got it. It is a bit weird that we cannot see apply
's signature until we dig deeper, but okay. Maybe leave a comment that briefly explains how weights
and ins
are passed through apply
to __call__
.
No description provided.