Quantization of Transformers model

[fairp]

Which is the best way to quantize a sentence transformer model in your opinion. I am referring to multilingual bert-based models.