Patrick Kidger

Hmm. That's definitely unfortunate, when transferring over pretrained models. This is probably down to `jax.lax.conv_general_dilated`, which `eqx.nn.Conv2d` is ultimately just a thin wrapper for. If you can, I'd suggest trying...

Great. Hopefully this can be resolved one way or another; I can imagine this would be a bit of a blocker for eqxvision. I'll leave this issue open for now.

Hey there - I'm afraid that's from a project that hasn't been released yet. Give it a few months and watch this space!

I'm not sure what you mean about `?` - I assume this is in reference to some piece of tooling you use? In any case, I'd be happy to include...

Ah, this will probably be because it's grabbing the metaclass `__call__` rather than using the class `__init__`. (I suppose `help` must have some special-casing internally to not exhibit the same...

With #133 we now have support for discrete terminating events! This provides support for steady-state finding, which is one of the big-ticket items for events.

These aren't currently available but agreed that this would make sense as a statistic to collect. This is the kind of thing that'd be really easy to do once (or...

I'm not a Flax user, so take this with a pinch of salt. But probably something like the following. ```python variables = model.init(...) def vector_field(t, y, args): return model.apply(args, y)...

I'm afraid not. This looks like something to do with Flax, which I'm not familiar enough with that I can help debug this. That said I'm sure it's possible. Both...

Hurrah! I'm glad you figured this out.