Abhirath Anand

Does Diffractor already work with most Flux models (or at least those with built-in layers)? I was under the impression that it wasn't there yet 😅

Speaking of Lux's documentation, it looks absolutely beautiful - any chance Flux would consider using a different theme?

> Again, what's the benefit beyond visually making the model simpler? I'll try and benchmark to see if there's a difference. But among other things, it makes porting weights from...

Another optimisation that caught my eye was the flattening of nested `Chain`s. Would that be something that would maybe help with TTFG and the backward pass times? (not sure if...

Could this be somehow extended to gradient errors as well? Zygote spits out _enormous_ stack traces, and by the time I find the error it's buried under an absolute mountain...

I'd also like to add one thing about the current docs structure that strikes me as a little unintuitive - the API reference is not grouped together but scattered through...

Is there a way to get Functors to only "see" down to a certain level? If `fmap` can somehow be overloaded to stop at the Flux layer level (for custom...

That's great! I tried something that's a pretty typical usecase and it worked quite well: ```julia julia> is_layer_or_leaf(m) = Functors.isleaf(m) is_layer_or_leaf (generic function with 1 method) julia> is_layer_or_leaf(::Conv) = true...

I was trying to give this another go, but I noticed the above example (from [here](https://github.com/FluxML/Flux.jl/issues/1879#issuecomment-1107684007)) doesn't work with DenseNet. The error was quite cryptic: ```julia julia> model = DenseNet();...

I am running into this issue while trying to implement DenseNet. Since `vcat` is one of the only non-mutating ways to append elements to arrays, this is a blocker for...