search

FrancescoSaverioZuppichini / glasses

High-quality Neural Networks for Computer Vision 😎
https://francescosaveriozuppichini.github.io/glasses-webapp/
MIT License
436 stars 37 forks source link

Feature/core #187 #188

Closed FrancescoSaverioZuppichini closed 3 years ago

FrancescoSaverioZuppichini commented 3 years ago

Done ViT -> why? What am I doing here? Reality is poison

codecov-io commented 3 years ago

Codecov Report

Merging #188 (945d07c) into develop (75fbc9e) will decrease coverage by 0.26%. The diff coverage is n/a.

Impacted file tree graph

@@             Coverage Diff             @@
##           develop     #188      +/-   ##
===========================================
- Coverage    97.81%   97.54%   -0.27%     
===========================================
  Files           58       63       +5     
  Lines         2109     2652     +543     
===========================================
+ Hits          2063     2587     +524     
- Misses          46       65      +19
Impacted Files Coverage Δ
glasses/nn/blocks/__init__.py 94.91% <0.00%> (-5.09%) :arrow_down:
test/test_VGG.py 100.00% <0.00%> (ø)
test/test_att.py 100.00% <0.00%> (ø)
test/test_UNet.py 100.00% <0.00%> (ø)
test/test_auto.py 100.00% <0.00%> (ø)
test/test_ResNet.py 100.00% <0.00%> (ø)
test/test_FishNet.py 100.00% <0.00%> (ø)
test/test_Storage.py 100.00% <0.00%> (ø)
glasses/nn/__init__.py 100.00% <0.00%> (ø)
test/test_protocols.py 100.00% <0.00%> (ø)
... and 59 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 2647f18...945d07c. Read the comment docs.

FrancescoSaverioZuppichini commented 3 years ago

We need to have .widths in the Encoder, would it be better to subclass the TransformerEncoder and create a ViTTransformerEncoder and add there the .widths so the vanilla TransformerEncoder is nice and clean