Multilingual (50 languages) embeddings using pytorch

[pankajarm]

Here is Multilingual (50 languages) embedding microservices using pytorch + Sentence Transformers library from https://www.sbert.net/index.html

SBERT When compare on various downstream tasks, such as in-domain and Zero-shot retrieval performances on BEIR dataset produces better results on embedding generated from tensorflow hub Universal Sentence Encoder (USE-QA).

Here is the research paper https://arxiv.org/pdf/2104.08663v1.pdf

I follow the exact same structure from tensorflow based embedding micro-services by @Ben-Epstein and just updates few line in main.py and requirements

you can check it it out multilingual embedding example, by installing requirements, running main.py and going to http://localhost:8088/docs