The main SentenceTransformer class has config_kwargs.

config_kwargs (Dict[str, Any], optional) – Additional model configuration parameters to be passed to the Hugging Face Transformers config. See the AutoConfig.from_pretrained documentation for more details.

This is not available through the SentenceTransformer Embedders in haystack (though we have model_kwargs and tokenizer_kwargs).

It could be good to also expose config_kwargs.

For instance, turning off use_memory_efficient_attention to run dunzhang/stella_en_400M_v5 seems to be not possible when loading through SentenceTransformersDocumentEmbedder.