Show HN: Comparing various contrastive losses on text and vision embeddingsgithub.com/viig99 3kartoolOz2ydiscuss