-
I brought up this as an issue in the `Zygote.jl` repository, but it might belong here:
`Zygote` fails to use `rrule`s defined by `TaylorDiff` when run with `Julia 1.10+`
In `Julia 1.9.4`:
```…
-
I guess half-precision is not officially supported by Flux yet, but as it "almost" works now, perhaps it is worth looking into:
```julia
julia> cc = Flux.paramtype(Float16, Conv((3,3), 3 => 64)) |…
-
The following throws an error on Zygote 0.5.10, but segfaults on Zygote 0.5.11 & beyond (with Julia 1.5.3 / ChainRules v0.7.33)
```julia
julia> using Zygote, StaticArrays, LinearAlgebra
julia> …
-
## Long rambling introduction
Distributions.jl currently allows many weird edge-cases which result in Dirac-type degenerate point mass distributions.
There are numerous bugs related to many of t…
-
`Zygote` fails to use `rrule`s defined by other packages when run with `Julia 1.10+`
I noticed this when looking at compatibility with [`TaylorDiff.jl`](https://github.com/JuliaDiff/TaylorDiff.jl).…
-
`gather` works fine on Apple Silicon, but with `scatter` I get an error.
The [scatter kernel](https://github.com/FluxML/NNlib.jl/blob/master/src/scatter.jl) works fine with cuda and amdgpu arrays.
…
-
This used to work with Zygote 0.3.1 but now gives the following compiler error for 0.3.2.
```
julia> function s(x)
t = 0.0
sign = -1.0
for i in 1…
-
```julia
julia> using Zygote
julia> function f(x)
d = Dict(1=>x)
d2 = copy(d)
return d2[1]
end
f (generic function with 1 method)
julia> gradient(f, 1)
…
-
Zygote doesn't interact too nicely with LazyArrays.jl it seems, e.g.:
```julia
julia> f(x) = sum(BroadcastArray(exp, x))
f (generic function with 1 method)
julia> Zygote.gradient(f, randn(10))
…
-
When working on packages that involve compiler internals, it's often useful to have Revise open and running for ease of experimentation, but also need to constantly kill the Julia session due to cachi…