Niklas Heim

Reverse over ForwardDiff differentiation: Zygote vs. ReverseDiff

5

I would like to compute the gradient of a function that contains a `ForwardDiff.gradient`/`ForwardDiff.derivative`. Computing that higher order gradient with ForwardDiff works, but will be slow for my case be...

bohrium.do_while much slower if loop function return np.any

3

from time import time import numpy as np def do_while(func, niters, *args, **kwargs): for i in range(niters): if not func(*args, **kwargs): break import bohrium as np do_while = np.do_while def...

First, thanks a lot for `Weave.jl`! I really like it and use it on a regular basis! :) ## description Running the `examples/julia_latex.texw` to produce a tex file produces incorrect...

Gradients of logpdf with TuringDiagMvNormal return nothing

3

This puzzles me a bit ```julia using DistributionsAD, Distributions, Flux using DistributionsAD: TuringDiagMvNormal Flux.@functor TuringDiagMvNormal m = [1.0] S = [0.1] f = TuringDiagMvNormal(m,S) x = 5 .+ randn(1,100) ps...

Once https://github.com/JuliaDiff/ChainRules.jl/issues/262 is fixed it would be nice to test gradients through `rand`.

Use rrule of KwFunc for Core.kwcall

2

This seems to fix non-differentiable keyword arguments by constructing the `KwFunc` defined in Diffractor.

More tests for Base

1

Includes the rest of the tests for Base (and moves the base tests to a separate file). More tests to come in the next PR!

Add tests for `fill_identities`. namely turn the following examples into tests: ```python b = chain(CNOT(0,1), CNOT(1,2)) b = chain(kron(CNOT(0,1)), kron(CNOT(1,2))) b = kron(CNOT(0,1)) b = chain(SWAP(0,1), SWAP(0,3)) ``` This could...

[Infra] Docker image for qadence installs

1

@dominikandreasseitz had the great idea of automatically building docker image with up-to-date qadence dependencies to speed up our CI pipelines. Let's do it!