Ruben

Just as a little bump, I'd also love to see electra support as I'm currently seeing the best results for my task (in german) with svalabs/cross-electra-ms-marco-german-uncased

> Same here. As noted in the Deepset [article](https://docs.cloud.deepset.ai/docs/language-models-in-deepset-cloud#ranker-models), in their practice, svalabs/cross-electra-ms-marco-german-uncased is the best model for Geman. If you havent already seen it, you could check out [michaelfeil/infinity](https://github.com/michaelfeil/infinity),...

It does at least for this case (or probably similar ones with very close to 1.0 scores) @michaelfeil TransformersSimilarityRanker Top 5 Scores: 1. 0.9997518658638 2. 0.9995856881141663 3. 0.999550998210907 4. 0.9986598491668701...

@michaelfeil Ok I've experimented a bit today: The issue does not come from the pytorch.sigmoid/np difference. Even though the results do not get rounded up to 1.0 using sigmoid, they...

Yeah, with float32 the scores are identical, thanks :)

Hey @julian-risch, For the time being it's just two but we're planning to support more in the future for different projects so that they can easily integrate the API. We're...

Ok as a little update, I've found a decent solution. Every heavier aspect of the initialization for my indexing pipelines now happens inside my fastapi background tasks. This way it...

Bumping this because it got force pushed somewhere but nothing happened after. Would be really nice, with the current darkmode I cant use hoppscotch, kinda hard on my eyes.

@stevkan The client is Teams. Kind of yeah, it's a RAG system with the possibility of images or other attachments being relevant to the answer. Attaching images (as additional attachments...

Have you found a solution @lord-maestro-encora ? The issue exists for me aswell, updated to 4.23.1 but it persists Since I cant find a good reason why it wouldnt work,...