korean-sbert-korsts

Korean Sentence-BERT finetuned on KorSTS dataset.

Performance

Metric	Baseline	Finetuned	Improvement
Spearman	0.6093	0.8101	+0.2009
Pearson	-	0.8147	-

Usage

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("seongyeon1/korean-sbert-korsts")
embeddings = model.encode(["문장 1", "문장 2"])

Training

Base: klue/bert-base
Loss: CosineSimilarityLoss
Epochs: 4
Batch: 32
LR: 2e-05

Downloads last month: -

Safetensors

Model size

0.1B params

Tensor type

F32