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.

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

Rational-quadratic spline transformations #15

Draft
wants to merge 14 commits into
base: main
Choose a base branch
from

Conversation

VasylHafych
Copy link
Member

Hi @oschulz, @Micki-D,

Here is a draft of the Monotonic Rational-quadratic Splines used in the Neural Spline Flows paper. Pullbacks for gradient evaluations are already implemented, but yet without @atomic operations. There is a demo notebook that allows playing with toy fits.

I will keep this WIP branch here to get your comments. Also, the updates with atomic operations should be pushed here soon.

@oschulz
Copy link
Member

oschulz commented Jun 28, 2022

Thanks! @Micki-D want to start playing with this? @VasylHafych , should be wait for the atomic ops before merging?

@VasylHafych
Copy link
Member Author

Yes, let's merge and review the final version.

@Micki-D
Copy link

Micki-D commented Jun 29, 2022 via email

@codecov-commenter
Copy link

codecov-commenter commented Jul 11, 2022

Codecov Report

Merging #15 (1f4b56a) into main (314aae1) will decrease coverage by 32.04%.
The diff coverage is 0.00%.

@@             Coverage Diff             @@
##             main      #15       +/-   ##
===========================================
- Coverage   66.37%   34.32%   -32.05%     
===========================================
  Files           6        7        +1     
  Lines         226      437      +211     
===========================================
  Hits          150      150               
- Misses         76      287      +211     
Impacted Files Coverage Δ
src/optimize_whitening.jl 0.00% <0.00%> (ø)
src/spline_trafo.jl 0.00% <0.00%> (ø)

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 314aae1...1f4b56a. Read the comment docs.

Remove @const macros

Fix array indexing in Kernels

Fix missing B arg

Fix bugs, seperate param and param grad evaluation

Fix example notebook
VasylHafych and others added 5 commits July 14, 2022 19:33
Pull out Array allocation from kernels
Separate parameter evaluation and parameter and parameter gradient evaluation
Remove Loops from pb_kernel(), rename utils, represent LogJac as Matr…
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants