SelfSupervisedLoss with a miner?

[ManuelZ]

Is it possible to use SelfSupervisedLoss with a miner? I'm wrapping CircleLoss.

[KevinMusgrave]

Unfortunately no. But it's a pretty simple wrapper, so the best workaround for now is to copy/paste the label-creating logic and not use the wrapper.

pytorch-metric-learning/src/pytorch_metric_learning/losses/self_supervised_loss.py

Lines 57 to 67 in 28d5acc

	labels = torch.arange(embeddings.shape[0])
	if self.symmetric:
	embeddings = torch.cat([embeddings, ref_emb], dim=0)
	labels = torch.cat([labels, labels], dim=0)
	return self.loss(embeddings, labels)
	return self.loss(
	embeddings=embeddings,
	labels=labels,
	ref_emb=ref_emb,
	ref_labels=labels.clone(),
	)