-
When the VJP is not an abstract array, things get weird
```julia
julia> import AbstractDifferentiation as AD
julia> import Zygote
julia> ad_backend = AD.ReverseRuleConfigBackend(Zygote.Zygot…
-
Hello,
A collegue pointed me to your package and it seems like it might be a useful method for speeding up parameter estimation for some of our models. Thank you for putting this together.
I h…
-
Take the following MWE for the custom rule of a `make_solution` rrule. NOte that it returns a SciMLBase class.
```julia
using SciMLBase, Zygote, ChainRulesTestUtils, ChainRulesCore
struct MySol…
-
ModelingToolkit allows to index into ODESolution via a symbol. However, currently, this causes problems with Optimization using Zygote gradients.
I tried working on the issue but need to learn more…
bgctw updated
3 years ago
-
On the Python side we have this lecture that motivates use of Python:
https://lectures.quantecon.org/py/about_py.html
It's about why rather than how. Perhaps the Julia side could benefit from s…
jstac updated
4 years ago
-
I'm comparing performance of the built-in Dense layer with an equivalent Linear layer and a Dist layer. There are a couple of problems.
Full code, output, Manifest, Project: https://gist.github.com…
-
I'm experiencing some slower performances when taking gradients on `CuArray`s, than normal CPU `Array`s, using Zygote. Here's an MWE (the Tullio operations implement a mixture linear regression):
```…
-
Currently, these distributions are broken due to https://github.com/FluxML/Zygote.jl/issues/873. As discussed in https://github.com/TuringLang/Bijectors.jl/pull/155, the corresponding tests are marked…
-
I have a matrix that I want to fill with 0 values using PaddedViews.jl:
`f2 = x -> sum(PaddedView(eltype(x)(0.0), x, (5, 5)))`
But when using zygote to calculate the gradient:
```
arr = collect(…
-
I am trying to implement a particular flavor of a variational autoencoder---the Riemannian Hamiltonian VAE from [this publication](http://arxiv.org/abs/2010.11518)---whose loss function involves two s…