[Feature] Implement I-JEPA

[crypdick]

Describe the feature

There have been a few joint embedding-based works lately that are getting great results without the need for hand-crafted data augmentations. In particular, I-JEPA seems to have great results, out-performing masked autoencoders.

The official reference implementation is here: https://github.com/facebookresearch/ijepa

Will you implement it?

• [ ] I would like to implement this feature and create a PR!

[lxysl]

Hi, I am willing to learn the I-JEPA code and try to implement it in MMPretrain.