Hi, we have multiple tasks that use SBERT sentence embedding, therefore we embed the text once and cache the sentence embedding vectors, then re-use the vectors everywhere. We found it really improve the time performance of our applications.
Is there a way for the model to take sentence embedding vectors as the input instead of text?
The following is how we use the model. Thank you
model = SBertSummarizer('all-mpnet-base-v2')
sum_3sent = model(text_prefer_sentence_vectors, num_sentences=3)
Hi, we have multiple tasks that use SBERT sentence embedding, therefore we embed the text once and cache the sentence embedding vectors, then re-use the vectors everywhere. We found it really improve the time performance of our applications.
Is there a way for the model to take sentence embedding vectors as the input instead of text? The following is how we use the model. Thank you
model = SBertSummarizer('all-mpnet-base-v2') sum_3sent = model(text_prefer_sentence_vectors, num_sentences=3)