Lightweight multilingual sentence encoder optimized for intent classification.
Created from paraphrase-multilingual-MiniLM-L12-v2 via layer pruning + corpus-based vocabulary pruning.
1from sentence_transformers import SentenceTransformer
2
3model = SentenceTransformer("L4_uniform")
4
5sentences = [
6 "예약 좀 해줘", # Korean
7 "What did I order?", # English
8 "今日はいい天気ですね", # Japanese
9 "Reserva una mesa", # Spanish
10]
11
12embeddings = model.encode(sentences)
13print(embeddings.shape) # (4, 384)
A distilled version of this model is also available with improved performance.
