Loss Scale for Training Siglip

Hi,

Thanks for your great work. I was trying to apply siglip loss for training contrastive models. However, I find the loss scale is quiet small, usually around 0.003 at the begging. I wonder if any thing goes wrong in my implementation.

```def sigclip_loss(logits):
n = logits.size(0)
labels = 2 * torch.eye(n) - torch.ones(n, n) # -1 with diagonal 1
labels = labels.to(logits.device)
loss = -torch.mean(F.logsigmoid(labels * logits)) / n
return loss```

No repository file or test is named. Start by reviewing the posted sigclip_loss implementation and comparing its normalization and expected scale with the repository's SigLIP training path. Done should establish whether the reported loss is expected or identify the specific implementation issue.

Written by the indexing model from the issue text.