Andrey Vasnetsov

there are 2 options I can see: - query multiple vectors with same filtering request - pros: simpler query, less overhead - cons: can't specify multiple parameters - allow specify...

https://github.com/qdrant/qdrant/issues/186

We are preparing this: https://ui.qdrant.tech/ I hope it will be even a little more convenient

the link should be displayed with proper version argument as a part of this PR: https://github.com/qdrant/qdrant/pull/681

@IvanPleshkov can you please share benchmarks result here?

> Are we also using flush(false) somewhere in the unit tests for sanity check? it is used as a part of optimizer. But I think writing an explicit test should...

Hi @keotic, take a look at https://qdrant.tech/documentation/search/#recommendation-api - it might be the thing you are looking for

If you are specifying just one ID - it will be similar to search. If you want to search for multiple stored vectors - you would need, indeed, a batch...

Looks like RAM and Disk usage statistics in API is more complicated to implement than I expected. Suggest to just remove it and rely on external tools

trying to compare xbm_loss and regular loss: ``` torch.random.manual_seed(42) embeddings = torch.rand(64, 100) groups = torch.randint(0, 10, (64,)) loss = TripletLoss(margin=1.0, mining="hard") print(loss(embeddings, groups)) print(loss._compute_xbm_loss(embeddings, groups, embeddings, groups)) ``` If...