search

lightly-ai / lightly

A python library for self-supervised learning on images.
https://docs.lightly.ai/self-supervised-learning/
MIT License
2.89k stars 249 forks source link

LIG-5222 Allow skipping weight initialisation for Masked Vision Transformers #1581

Closed ersi-lightly closed 1 month ago

ersi-lightly commented 1 month ago

Addresses issue #1511 and allows to skip weight initialization for MaskedVisionTransformerTIMM, MaskedVisionTransformerTorchvision and MAEDecoderTIMM. This will enable to load a pretrained vit model.

codecov[bot] commented 1 month ago

Codecov Report

Attention: Patch coverage is 0% with 6 lines in your changes missing coverage. Please review.

Project coverage is 81.77%. Comparing base (d09a6b5) to head (a60d5b3).

Files Patch % Lines
lightly/models/modules/masked_autoencoder_timm.py 0.00% 2 Missing :warning:
...y/models/modules/masked_vision_transformer_timm.py 0.00% 2 Missing :warning:
...s/modules/masked_vision_transformer_torchvision.py 0.00% 2 Missing :warning:
Additional details and impacted files ```diff @@ Coverage Diff @@ ## master #1581 +/- ## ========================================== - Coverage 81.81% 81.77% -0.05% ========================================== Files 144 144 Lines 6120 6123 +3 ========================================== Hits 5007 5007 - Misses 1113 1116 +3 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.