Generalizations

[mloubout]

PR covers following points:

• Removed all slim layers and networks as discussed in meeting
• Make all layers shape independent (not more nx/ny/nz). Only exception is ConditionalResidualBlock that has parameters that depend on the shape
• Added basic Chanrules implementation

Need to check all examples are still running

[codecov[bot]]

Codecov Report

Merging #22 (5657ad6) into master (ee8c395) will increase coverage by 4.89%. The diff coverage is 79.06%.

@@            Coverage Diff             @@
##           master      #22      +/-   ##
==========================================
+ Coverage   72.85%   77.75%   +4.89%     
==========================================
  Files          34       31       -3     
  Lines        2932     2418     -514     
==========================================
- Hits         2136     1880     -256     
+ Misses        796      538     -258     

Impacted Files	Coverage Δ
src/utils/neuralnet.jl	65.51% <0.00%> (-0.77%)	:arrow_down:
src/layers/invertible_layer_glow.jl	86.90% <33.33%> (-1.05%)	:arrow_down:
...rc/networks/invertible_network_conditional_hint.jl	80.53% <33.33%> (-0.72%)	:arrow_down:
.../invertible_network_conditional_hint_multiscale.jl	63.15% <33.33%> (-0.38%)	:arrow_down:
src/networks/invertible_network_glow.jl	79.31% <33.33%> (-0.69%)	:arrow_down:
src/networks/invertible_network_hint_multiscale.jl	67.50% <33.33%> (-0.57%)	:arrow_down:
src/layers/invertible_layer_basic.jl	86.11% <50.00%> (-0.38%)	:arrow_down:
src/networks/invertible_network_irim.jl	69.87% <50.00%> (+9.87%)	:arrow_up:
src/InvertibleNetworks.jl	55.55% <57.14%> (+35.55%)	:arrow_up:
src/conditional_layers/conditional_layer_hint.jl	87.50% <60.00%> (+3.69%)	:arrow_up:
... and 11 more

---

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 ee8c395...5657ad6. Read the comment docs.