search

slimgroup / InvertibleNetworks.jl

A Julia framework for invertible neural networks
MIT License
149 stars 21 forks source link

add split_scales option to glow #40

Closed rafaelorozco closed 2 years ago

rafaelorozco commented 2 years ago

Giving Glow the option to not be multiscale. Since this turns off squeezes this allows for input of dimensions such as (1,1,nc,batchsize).

This gives a simple-to-use architecture option for non conditional NF on simple 2d gaussian/rosenbrock examples.

codecov[bot] commented 2 years ago

Codecov Report

Merging #40 (869798a) into master (0332c97) will increase coverage by 1.35%. The diff coverage is 100.00%.

Impacted file tree graph

@@            Coverage Diff             @@
##           master      #40      +/-   ##
==========================================
+ Coverage   85.14%   86.50%   +1.35%     
==========================================
  Files          31       31              
  Lines        2451     2460       +9     
==========================================
+ Hits         2087     2128      +41     
+ Misses        364      332      -32
Impacted Files Coverage Δ
.../invertible_network_conditional_hint_multiscale.jl 87.41% <100.00%> (+10.59%) :arrow_up:
src/networks/invertible_network_glow.jl 99.09% <100.00%> (+9.98%) :arrow_up:
src/utils/dimensionality_operations.jl 94.33% <100.00%> (+3.77%) :arrow_up:

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 0332c97...869798a. Read the comment docs.