guarin

@ishaanagw I unassigned you from the issue to make it clearer that other people can also work on it.

These models in `lightly/models` are deprecated: Regarding 2) you can just put a local `# type: ignore[error code]` comment where the `error code` is the reported error by mypy. See...

Hi! Multi-modal is definitely something we would like to incorporate. There are two main components missing for this: Data loading for text, and NLP models/tokenizers. For both cases we have...

Hi! > You want a more general abstract class for the decoder which can then be used for MAE and I-JEPA accordingly. Am i correct to understand this? Yes, the...

Thanks for the extra info! We don't strictly follow the factory based pattern. We have some plans to update the package structure but they are still under discussion. Our goal...

Hi, sorry for the late reply. It looks like your loss is way too large (2415920191337764664519950336.00000). Maybe try decreasing the learning rate or check your gradient values (clip them if...

Hi! You are correct that the focus of the library is on training models from scratch. We haven't incorporated finetuning because there are many possible finetuning tasks (classification, detection, segmentation,...

Ah sorry I didn't understand the question correctly! In principle everything described in the paper should already be possible with Lightly AFAIK (awesome paper btw!). If you have a pretrained...

This might be an interesting topic for a tutorial 🤔

Hi! For Grad-CAM it is probably best if you just use the grad cam library: https://github.com/jacobgil/pytorch-grad-cam For the attention map visualization from DINO ([here](https://github.com/facebookresearch/dino/blob/main/visualize_attention.py)) the issue is mostly on how...