AbstractDifferentiation.jl
AbstractDifferentiation.jl copied to clipboard
JuliaDiff
An abstract interface for automatic differentiation.
Bumps [julia-actions/cache](https://github.com/julia-actions/cache) from 1 to 2. Release notes Sourced from julia-actions/cache's releases. v2.0.0 Breaking Changes ⚠️ v2.0.0 requires node20. This is a breaking change, because node20 does not support the...
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot] avatar"){kind=avatar}
Bumps [julia-actions/setup-julia](https://github.com/julia-actions/setup-julia) from 1 to 2. Release notes Sourced from julia-actions/setup-julia's releases. v2.0.0 - Update to Node20 What's Changed update to node20 by @ranocha in julia-actions/setup-julia#209 Note the rationale for...
Hi there, Adrian Hill (@adrhill) and I recently started a new unified differentiation package called DifferentiationInterface.jl: > https://github.com/gdalle/DifferentiationInterface.jl I'm gonna try to show why we did that, how we proceeded,...
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.ZygoteRuleConfig()) AbstractDifferentiation.ReverseRuleConfigBackend{Zygote.ZygoteRuleConfig{Zygote.Context{false}}}(Zygote.ZygoteRuleConfig{Zygote.Context{false}}(Zygote.Context{false}(nothing))) julia> AD.second_derivative(ad_backend, identity, 1) ERROR:...
One of the big downsides of AbstractDifferentiation.jl is the heavy use of (sometimes nested) closures and `if`-based dispatch, which generates type instability. I think many of those are fixable, in...
When I bumped AbstractDifferentiation compat to 0.6 on ImplicitDifferentiation, I got type inference errors in ChainRulesTestUtils that were not there before. Probably a regression due to the latest changes. Any...
I was chatting with @adrhill and he suggested that the macro `@primitive` could be discarded if each backend simply implemented some methods from AbstractDifferentiation, mostly `jacobian` and a `pushforward` or...
All packages made for performance (ForwardDiff.jl, FiniteDiff.jl, SparseDiffTools.jl), include some kind of caching interface. For example, instead of `ForwardDiff.jacobian(f,x)`, you should call `ForwardDiff.jacobian(f,x,config)`. `config`, `cache`, etc. is all important for...
This is based on https://github.com/julia-actions/julia-invalidations. Adding such checks came up in https://discourse.julialang.org/t/potential-performance-regressions-in-julia-1-8-for-special-un-precompiled-type-dispatches-and-how-to-fix-them/86359. I suggest to add this check here since this package is widely used as a dependency. See also...
Bumps [codecov/codecov-action](https://github.com/codecov/codecov-action) from 3 to 4. Release notes Sourced from codecov/codecov-action's releases. v4.0.0 v4 of the Codecov Action uses the CLI as the underlying upload. The CLI has helped to...