Will the trained segmentation head be released?

Thank you for sharing your excellent work and the PE checkpoints!

I saw in the paper (Appendix B.3.3) that you evaluated semantic segmentation on ADE20K using a linear + convolutional probe on frozen PE features. I was wondering if you plan to release the trained segmentation head (e.g., weights or code for the dense prediction probe)?

It would be really helpful for reproduction and downstream use. Thanks again for the contribution!