guarin
guarin
Does the same happen if you don't load the lightly pre-trained checkpoint? My suspicion would be that the learning rate is too high. Some other points to consider: * For...
Could you provide a minimal reproducible example so we can test on our side?
Looking at your script you should first try to increase the batch size. SimCLR works best with large batch sizes. I would also suggest to reduce temperature to `0.1`.
I'll close this issue for now, feel free to reopen if you have more questions :)
Thanks a lot for the PR! The review might take a while due to the upcoming holiday season.
Hi! We have a tutorial on how to use timm models with Lightly: https://docs.lightly.ai/self-supervised-learning/tutorials/package/tutorial_timm_backbone.html We don't have weights for MSN available
Hi! I suggest to use TorchGeo for multispectral data. They have a example using SimCLR here: https://github.com/microsoft/torchgeo/blob/main/torchgeo/trainers/simclr.py I believe it is only needed for data loading/augmentations. The actual model implementation...
This looks interesting, I haven't encountered this before. What type of data are you using? And do you use sync batchnorm?
Hi @ishaanagw! Thanks for contributing, feel free to work on any file :) > Hey @guarin, some of the mypy tests are failing in local, 48 errors in 23 files...
> Hey @guarin, some of the mypy tests are failing in local, 48 errors in 23 files in master, do you think I can still push my changes for the...