vprelovac/universal-sentence-encoder-large-5

This is a part of the MTEB test.

# !pip install tensorflow_text 

import tensorflow_hub as hub
from tensorflow_text import SentencepieceTokenizer
import tensorflow as tf

embedder=hub.load("https://tfhub.dev/google/universal-sentence-encoder-multilingual-large/3")

class USE():
    def encode(self, sentences, batch_size=32, **kwargs):
        embeddings = []
        for i in range(0, len(sentences), batch_size):
            batch_sentences = sentences[i:i+batch_size]
            batch_embeddings = embedder(batch_sentences)
            embeddings.extend(batch_embeddings)
        return embeddings


model = USE()