Add TFData2VecVision for semantic segmentation #17271
Rocketknight1
merged
21 commits into
huggingface:main
from
tf-data2vec-vision-seg
feat: initial implementation of data2vec segmentation model in TF.
4b421f16
chore: minor corrections to make the segmenter work.
6c5a5098
chore: removed unncessary files.
1bf19a3b
chore: add tests and other modifications.
f538a638
fix: loss computation for segmentation.
857fc5fc
chore: remove unused variable.
5f5718e8
chore: formatting.
6e601c2e
sgugger
approved these changes
on 2022-05-16
added a dummy adaptive pooling layer.
b939a9d8
removed unnecessary file.
f9292cf2
potentially add identifiers to layer names.
8ccf88bf
fix: layer naming.
63b7e464
chore: removed unnecessary print.
6d027603
Merge pull request #1 from sayakpaul/fix/tf-data2vec-seg
b4a49ad8
Skipping unneeded test
d7d375c6
chore: add logging to debug tolerance.
38127e4f
fix: segmentation tests for tfdata2vecvision
a9bbc8f1
Merge branch 'main' into tf-data2vec-vision-seg
4d2cef77
chore: make style.
4c61e4fc
fix: layer names, assertion to be resolved.
1b2d5476
Bumping test tolerance a bit
be835373
chore: bump the tol in PT test.
ef65922a
Rocketknight1
approved these changes
on 2022-06-08
Login to write a write a comment.
Login via GitHub