Have you provide the code for finetuning Siglip2 using all the loss (captioning, dense captioning, self-distillation, masked prediciton) as mentioned in the original paper?

Have you provide the code for finetuning Siglip2 using all the loss (captioning, dense captioning, self-distillation, masked prediciton) as mentioned in the original paper?

No file, entry point, or test is named. Start by comparing the repository's existing SigLIP2 finetuning support with the original paper's captioning, dense captioning, self-distillation, and masked prediction losses. Done would require code covering all four losses, with a clear way to verify each one.

Written by the indexing model from the issue text.