Nicolas Patry

Not in the very near future. However we could build bindings around https://www.amd.com/en/graphics/servers-solutions-rocm first in order to start working. I'm not too familiar with the workflow, I do know there...

@bayedieng could you try out : https://github.com/huggingface/candle/pull/1523 maybe ? It's possible for metal to be slower if there's not enough memory available to run the model I think, otherwise it...

> because both phi 2 and stable diffusion turbo ran faster on metal then accelerate in the past. The SIMD line is this single one: https://github.com/huggingface/candle/blob/main/candle-metal-kernels/src/lib.rs#L1277 Otherwise it's because of...

Multiple issues things at play that are currently known about: - Model loading is not really using correct information. `api-inference` doesn't know how to "guess" the model size properly, so...

Or the OpenAPI spec. Since community don't have too many parameters and have regular output types it could be the easiers (the transformers API was a nightmare to maintain exhaustively...

Isn't there a setting or a config that could be used within `sentence-transformers` to know which similarity to use ? It feels like this shouldn't be passed by users as...

I also remember there was such a configuration, but while writing this pipeline, we were told that cos_sim had basically won. (FYI)

I don't have a super strong opinion against making a parameter but I have multiple indicators that make me feel like it's not correct. - The widgets don't send parameters...

IMO `generic` is not meant to be used very much. If it's tedious to use it's OK. AFAIK, `generic` is just meant to be used as a demo purpose for...

No there's not, truncating is the only simple option. The reason is that the API for non TGI models use the pipeline, which raises an exception that would need to...