Skip to content

Upper/LowerTriangular gradients #9

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

Jump to bottom
Open
wants to merge 8 commits into
base: master
Choose a base branch
from
Open

Upper/LowerTriangular gradients #9

wants to merge 8 commits into from

Conversation

pshashk
Copy link
Contributor

@pshashk pshashk commented Mar 26, 2019

No description provided.

@MikeInnes
Copy link
Member

Looks good to me, but can you do a quick rebase / merge master so that the tests pass?

@pshashk
Copy link
Contributor Author

pshashk commented Apr 4, 2019

Somehow *Triangular gradient definitions overwrite each other and UpperTriangular(::TrackedArray) returns tracked LowerTriangular. Minimal example with current master:

using LinearAlgebra, Tracker
using Tracker: track, data, @grad
using LinearAlgebra: UpperTriangular, LowerTriangular

LinearAlgebra.UpperTriangular(a::TrackedMatrix) = track(UpperTriangular, a)
LinearAlgebra.LowerTriangular(a::TrackedMatrix) = track(LowerTriangular, a)
@grad UpperTriangular(a) = UpperTriangular(data(a)), Δ -> (UpperTriangular(Δ),)
@grad LowerTriangular(a) = LowerTriangular(data(a)), Δ -> (LowerTriangular(Δ),)

UpperTriangular(param(rand(4,4)))
Tracked 4×4 LowerTriangular{Float64,Array{Float64,2}}:
 0.938394   ⋅         ⋅         ⋅      
 0.755833  0.960363   ⋅         ⋅      
 0.426979  0.843864  0.238508   ⋅      
 0.101401  0.10873   0.555092  0.381779

@mohamed82008 mohamed82008 mentioned this pull request Mar 13, 2020
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@pshashk @MikeInnes