Mike J Innes
Mike J Innes
@staticfloat would be great to get a once-over on this.
I don't think it's too difficult to do if you're willing to get your hands dirty; you'll want to look at [NCCL](https://github.com/vchuravy/NCCL.jl) and make copies of the model for each...
@tejank10 are you happy with the changes made here, or is there more to do?
It will also need to be in its own folder, and have a simple README. Otherwise this is looking good I think, but it'd be good to hear from @tejank10.
Cool! Hopefully you can use Flux's new ConvTranspose layer before too long. Also, can you open an issue on why you're not using the Flux training loop? I guess just...
I guess this model slipped through the net when we were upgrading to 1.0. A patch would be great if you're able to upgrade it.
Sure, feel free to do it as part of this PR.
Looks like the neural_odes.jl change is WIP? I went ahead and merged the lotka volterra changes as that's a really nice demo.
Aaaah, ok. I've actually used the additional noise term in a bunch of places, but hadn't at all connected this numerical issue to the lack of a noise term. You've...
Calling `precompile` on that function won't make a lot of difference because the compiler won't be able to infer the whole call chain. But you could try actually calling `stringmime`...