Skip to content
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.

Sign up for GitHub

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

Project nothing to nothing #1075

Closed
wants to merge 3 commits into from
Closed

Project nothing to nothing #1075

Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c53df99
NoTangent -> nothing
DhairyaLGandhi Sep 23, 2021
4034bed
add wrap_chainrules_output rule
DhairyaLGandhi Sep 24, 2021
7ad6e8b
typo
DhairyaLGandhi Sep 24, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions src/compiler/chainrules.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,6 +106,7 @@ Convert `x` from the differentials types ChainRules uses to the format Zygote us
# Zygote convention: even if many AbstractZero partials (i.e. multi-input function), make just 1 nothing.
@inline wrap_chainrules_output(x::Tuple{Vararg{ChainRules.AbstractZero}}) = nothing
@inline wrap_chainrules_output(x::ChainRules.AbstractZero) = nothing
@inlint wrap_chainrules_output(x::ChainRules.NoTangent) = nothing
Copy link
Member

@mcabbott mcabbott Sep 24, 2021 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This seems completely safe to me. Although might be covered by the AbstractZero line above?

Re PR's original change c53df99 to propagate nothings, I was surprised this broke no tests. I don't recall now what test motivated that method in the first place, and it's possible that some later change removed the need.

Does this solve SciML/SciMLSensitivity.jl#493? As noted there, I can't reproduce this locally.

for T_outer in (:Tuple, :NamedTuple)
# we create separate methods rather than using a `Union` + an `if` so that we avoid a
# branch that changes output type, because nested AD on that kinda thing makes Zygote less
Expand Down