New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Make sensitivities for structured matrix arguments structured #52
Comments
Since (say) frule(::typeof(*), ::Diagonal, ::AbstractVector) = nothing
rrule(::typeof(*), ::Diagonal, ::AbstractVector) = nothing Ref a similar question I posted in Zygote: FluxML/Zygote.jl#316 |
I think "disabling" the rules by returnuing nothing would do it, |
I think there are too many specializations in LinearAlgebra (and likely in many other in the wild). But maybe not so bad at least as a "midterm" solution? Looking at how Zygote handles broadcasting, I have a feeling that it'll take some time to support broadcasting properly... If ChainRules.jl's stance is "rules specializations are welcome" I can make a PR to implement FluxML/Zygote.jl#316 in ChainRules.jl (after the overhaul #91). |
yeah, I think bailing out sometimes is the right answer for ChainRules,
Yeah, that would be great. |
Quoting Will in #29:
The text was updated successfully, but these errors were encountered: