You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have been wanting to write something up on this for a while.
The kind of docs that are almost a blog post
This issue is to gather notes.
I think it is non-obvious that you do want to sometimes change the primal computation.
But you do sometimes, and in those times it is crucial that you can.
Its one of the features we have to support 1% of cases, which we really couldn't do as well if we didn't.
willtebbutt
changed the title
Design docs on Primal computations that do are not simply call the primal function
Design docs on Primal computations that do not simply call the primal function
Nov 20, 2020
I have been wanting to write something up on this for a while.
The kind of docs that are almost a blog post
This issue is to gather notes.
I think it is non-obvious that you do want to sometimes change the primal computation.
But you do sometimes, and in those times it is crucial that you can.
Its one of the features we have to support 1% of cases, which we really couldn't do as well if we didn't.
sincos
in rules for sin and cos ChainRules.jl#291\
and/
(Re)use factorizations in primal for \ etc in pullback ChainRules.jl#250getindex
Inplace getindex rrule ChainRules.jl#240The text was updated successfully, but these errors were encountered: