Kevin Musgrave

`enqueue_idx` is now `enqueue_mask` starting in v2.0.0. So a different number of embeddings can be enqueued by each process.

`losses.SelfSupervisedLoss` is now available in v2.0.0. The multiple reference embeddings aren't yet available, as I'm not sure what the best approach is for computing the loss. I've created an issue...

I've removed this loss function from v2.0.0 as I don't have time to figure out the bug. If someone figures it out, please open a pull request.

@pgrosjean Sorry for the delayed response. I've been extra busy the past few weeks. Thank you for investigating this and providing such a detailed explanation with plots! Yes, please open...

Too much work

Added by @domenicoMuscill0 in v2.4.0: ``` pip install pytorch-metric-learning==2.4.0 ``` https://kevinmusgrave.github.io/pytorch-metric-learning/losses/#rankedlistloss

Thanks @TKassis 😄 When the implementation was discussed (#372) it seemed like this loss would only ever be used as shown in Figure 1 of [the paper](https://arxiv.org/pdf/2105.04906.pdf):  So `embeddings`...

In v2.0.0 it is now `ref_includes_query`

Closing due to inactivity

Please do. I never got around to working on it.