Zygote.jl

Zygote.jl copied to clipboard

Reverse on reverse fails with a Flux nn

10

```julia using Flux, Zygote model = Dense(3,1) grad_f(x) = gradient(x -> sum(model(x)),x)[1] Zygote.jacobian(grad_f,rand(3)) ERROR: Can't differentiate foreigncall expression. You might want to check the Zygote limitations documentation. https://fluxml.ai/Zygote.jl/dev/limitations.html Stacktrace: [1]...

YichengDWu

`one'(x)` errors, although `gradient(one, 0.0) === nothing` does not

1

```julia using Zygote julia> one'(0.0) ERROR: MethodError: no method matching getindex(::Nothing, ::Int64) Stacktrace: [1] (::Zygote.var"#59#60"{typeof(one)})(x::Float64) @ Zygote ~/.julia/packages/Zygote/FPUm3/src/compiler/interface.jl:83 [2] top-level scope @ REPL[5]:1 ```

cossio

LoadError: Non-differentiable function Core.Intrinsics.lshr_int

2

Hi, here are some approaches that perform a sorted search. Each row in `x` should be searched for an the corresponding column of `s`. All `get_index` functions give the correct...

fedario

Directly call `ForwardDiff` to compute the hessian

16

Closes #1264 `Zygote.forward_jacobian` is not well maintained and the motivation for using it is not clear.

YichengDWu

second-order gradient throws foreigncall error

2

Zygote v0.6.41, Julia 1.7.3 MWE: ```julia using Zygote α, β = randn(2, 2), randn(2, 2) g(v) = map(eachcol(v), eachcol(β)) do x, y sum(x.*x.*y) end |> sum # this fails gradient(α)...

baedan

Taking nested gradient for implementing Wasserstein GAN with gradient penalty (WGAN-GP) on GPU

9

I am trying to implement WGAN-GP using Flux and Zygote. My implementation works fine on CPU but fails on GPU with error `LoadError: this intrinsic must be compiled to be...

bhatiaabhinav

`jacobian` is not differentiable

3

```julia julia> function f(x, bias) jac = Zygote.jacobian(x->x.^3, x)[1] return jac * x .+ bias end f (generic function with 1 method) julia> x,bias = rand(3),rand(3) ([0.2279638899624825, 0.6476786632858718, 0.13745627655377346], [0.051516386842686224,...

YichengDWu

Wrong Fill'ed gradient

1

```julia julia> Zygote.gradient(x -> sum(x .* randn.()), randn(5)) (Fill(-0.30782754826253156, 5),) ``` This is wrong. The result should be the random vector of coefficients.

cossio

Zygote, a FillArray of structs and broadcasting don't work together

2

Hi! Here is a MWE that gives an error when trying to differentiate a function that broadcasts a FillArray of structs. ```julia using FillArrays, Zygote struct T a end f(t::T,...

mohamed82008

Slow code and excessive allocations when backpropagating broadcasting

10

There seems to be a bug lurking somewhere in a recent update of Zygote and/or dependencies (I cannot really pinpoint it, but I noticed a significant slowdown of some code...

Sleort