Mike J Innes

Hey Sarvjeet, sorry for not getting back earlier on this. This looks like a good list of functionality. I'll try to review your PRs over the next couple of days.

I like your work so far. Can you submit a proposal? Would like to see some thoughts on HTTP.jl, as well as what you'd use as a test case. Perhaps...

Will take a look. I think it'd be good to build a small web app as a test case, (a) to prioritise features from the list you have above, and...

Thanks for the patch, glad to have this although I think it could be simplified a little (will leave a note).

It would be good to add `scheme` to the request dict as you've suggested, to allow routing on it. Unfortunately I have no idea how to get the `HttpServer.jl`'s URI...

I spent some time refactoring this stuff for exactly this reason. We now have [this version of update](https://github.com/FluxML/Flux.jl/blob/3a4c6274fadf1b94468da455a06f070b2cda6a64/src/optimise/train.jl#L9) which only requires a param -> grad mapping. It should be pretty...

The problem is that adapt can be called with anything, e.g. `adapt(CuArray, 2.0)`. In cases like that you don't want to do anything, since `2.0` is already GPU compatible (and...

Yeah, I think there's a small mindset difference going on here: in ML we tend to see the channels as a stack of related images, rather than as a single...

In theory this should be fairly simple, just a case of registering a C function pointer (created from a Julia function) with MathLink, which can accept some Mathematica expression objects...

Yeah, WSTP wasn't around at all when I wrote this and I'd bet money that it's just a trivial rename. Presumably they'll keep supporting the `ML*` functions as aliases in...