M-Chimiste/MiniLM-L-12-StackOverflow

Cross-Encoder for MS Marco

This model is a generic masked language model fine tuned on stack overflow data. It's base pre-trained model was the cross-encoder/ms-marco-MiniLM-L-12-v2 model.

The model can be used for creating vectors for search applications. It was trained to be used in conjunction with a knn search with OpenSearch for a pet project I've been working on. It's easiest to create document embeddings with the flair package as shown below.

Usage with Transformers

from flair.data import Sentence
from flair.embeddings import TransformerDocumentEmbeddings

sentence = Sentence("Text to be embedded.")
model = TransformerDocumentEmbeddings("model-name")
model.embed(sentence)
embeddings = sentence.embedding