Michael Abbott

Results 222 issues of Michael Abbott

This adds https://github.com/JuliaLang/julia/pull/39071 Note that: * ~~The version check may need to change once backported to 1.11 branch~~ * Before Julia 1.9, it won't use TwicePrecision, as the necessary log/exp...

The present readme is pretty awful. It should start by pointing you to the highest-quality examples we have, and say where to look to see various features demonstrated.

All models marked `☀️ v0.13` or `☀️ v0.13 +` should work fine with [Flux 0.14](https://github.com/FluxML/Flux.jl/releases/tag/v0.14.0). The only code change should be that, in some cases, `using CUDA` needs to be...

good first issue
update

This previously didn't have a manifest. ~~Needs a correction to the translation rule for Optimisers.ClipNorm in Flux.~~ fixed in https://github.com/FluxML/Flux.jl/pull/2145 Also, a simple test showed increasing loss, so perhaps something...

update

This is intended for https://github.com/FluxML/Fluxperimental.jl/pull/20 but probably a good idea anyway. Motivating example is something like this (where bigger arrays will print pages of numbers): ```julia julia> struct Tmp2; x;...

Aims to fix #2325 by mentioning that withgradient accepts a tuple. And to fix https://github.com/FluxML/Flux.jl/issues/2303 by moving discussion of old-style optimiser structs later. Also has training doc updates about L1,...

documentation

Uses https://github.com/FluxML/NNlib.jl/pull/457 to speed up & save memory, up to half the memory for a forward pass. Largest savings in the gradient will be for large batch size, and activation...

performance
run downstream test

Here it concludes that there are non-trainable arrays, but in fact the same weights appear in two layers, which throws off the counting: ```julia julia> using Flux julia> let d...

Like https://github.com/FluxML/Zygote.jl/pull/1281 this adds https://github.com/timholy/SnoopCompile.jl Timing some `gradient` of a model, it's not so obvious this saves time. M1 mac, Julia 1.9- nightly ``` ### Zygote v0.6.49, Flux v0.13.6 %...

Apparently @owainkenwayucl was trying out Flux, and @giordano was helping him out. https://twitter.com/owainkenway/status/1620771011863121921 https://github.com/owainkenwayucl/JuliaML/blob/main/Fashion/simple.jl Edit: now at https://www.youtube.com/watch?v=Yd1JkPljpbY It's useful to see what problems newcomers run into. Especially people new...

documentation