Michael Abbott

Agree the output shouldn't change, but you could easily reshape afterwards. But still not viable until `eachslice` takes multiple dims.

> equivalent to eachobs, something that cannot be obtained with batchsize=1 Isn't this `batchsize=-1`? That's what the docs seem to say: ``` julia> DataLoader(ones(2,3); batchsize=1) |> first 2×1 Matrix{Float64}: 1.0...

Perhaps ideally, `batchsize=0` would make more sense for the "individual obs without batch dim" command, and `batchsize=-1` for "wrap around to largest possible batch". Both are special cases. But this...

Yes I'm sure not materialising was the goal. However, some views are more useful than others. I wonder if the default should be something like `splitobs(ones(1,10); at=0.5)` makes two contiguous...

The use is functions like this, which load data & make two `DataLoader`s with the specified batch size: https://github.com/FluxML/model-zoo/blob/52420da6fcadf30ae2e190fc77669fe1d255ff10/vision/conv_mnist/conv_mnist.jl#L71-L84

You could almost use `typemax(Int)` for this purpose, apart from this warning: ```julia julia> DataLoader([1 2 3; 4 5 6]; batchsize=99, partial=false) |> collect ┌ Warning: Number of observations less...

These axes represent images the same size, with offset indices: ``` julia> length.((62:285, 18:241, 1:3)) (224, 224, 3) julia> length.((38:261, 18:241, 1:3)) (224, 224, 3) ``` I presume the previous...

Flux has its own definition https://github.com/FluxML/Flux.jl/blob/d4f1d816563edd5f953a3fd1ef7dd960d507ed22/src/layers/stateless.jl#L32 and issue https://github.com/FluxML/Flux.jl/issues/1952 about the name / location. Julia already uses "normalize" to mean something different to either of those: ``` help?> LinearAlgebra.normalize normalize(a,...

> `gradient(x -> sum(setindex!!(x, 2, 2)), zeros(3))` would not work with current BangBang since it just uses standard `setindex!`. Thanks for having a look! I guess that was my question...

Test failure on Julia 1.0: ``` Error During Test at /home/runner/work/DiffRules.jl/DiffRules.jl/test/runtests.jl:46 11 Test threw exception 12 Expression: isapprox(dy, finitediff((z->begin 13 SpecialFunctions.gamma(foo, z) 14 end), bar), rtol=0.05) 15 MethodError: no method...