albert-small-kor-sbert-v1.1

This is a sentence-transformers model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search.

albert-small-kor-v1 모델을 sentencebert로 만든 모델.

Usage (Sentence-Transformers)

Using this model becomes easy when you have sentence-transformers installed:

pip install -U sentence-transformers

Then you can use the model like this:

from sentence_transformers import SentenceTransformer
sentences = ["This is an example sentence", "Each sentence is converted"]

model = SentenceTransformer('bongsoo/albert-small-kor-sbert-v1.1')
embeddings = model.encode(sentences)
print(embeddings)

Usage (HuggingFace Transformers)

Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.

from transformers import AutoTokenizer, AutoModel
import torch


def cls_pooling(model_output, attention_mask):
    return model_output[0][:,0]


# Sentences we want sentence embeddings for
sentences = ['This is an example sentence', 'Each sentence is converted']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('bongsoo/albert-small-kor-sbert-v1.1')
model = AutoModel.from_pretrained('bongsoo/albert-small-kor-sbert-v1.1')

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling. In this case, cls pooling.
sentence_embeddings = cls_pooling(model_output, encoded_input['attention_mask'])

print("Sentence embeddings:")
print(sentence_embeddings)

Evaluation Results

• 성능 측정을 위한 말뭉치는, 아래 한국어 (kor), 영어(en) 평가 말뭉치를 이용함
  한국어 : korsts(1,379쌍문장) 와 klue-sts(519쌍문장)
  영어 : stsb_multi_mt(1,376쌍문장) 와 glue:stsb (1,500쌍문장)
• 성능 지표는 cosin.spearman
• 평가 측정 코드는 여기 참조

• 모델	korsts	klue-sts	glue(stsb)	stsb_multi_mt(en)
  distiluse-base-multilingual-cased-v2	0.7475	0.7855	0.8193	0.8075
  paraphrase-multilingual-mpnet-base-v2	0.8201	0.7993	0.8907	0.8682
  bongsoo/albert-small-kor-sbert-v1	0.8305	0.8588	0.8419	0.7965
  bongsoo/klue-sbert-v1.0	0.8529	0.8952	0.8813	0.8469
  bongsoo/kpf-sbert-v1.1	0.8750	0.8900	0.8863	0.8554
  bongsoo/albert-small-kor-sbert-v1.1	0.8526	0.8833	0.8484	0.8286

For an automated evaluation of this model, see the Sentence Embeddings Benchmark: https://seb.sbert.net

Training

• albert-small-kor-v1 모델을 sts(10)-distil(10) 훈련만 시킴 (nli-sts추가 훈련시키면 정합도가 떨어짐)
• 교사모델은 kpf-sbert-v1.1 이용함.

The model was trained with the parameters:

공통

• do_lower_case=1, correct_bios=0, polling_mode=cls

1.STS

• 말뭉치 : korsts(5,749) + kluestsV1.1(11,668) + stsb_multi_mt(5,749) + mteb/sickr-sts(9,927) + glue stsb(5,749) (총:38,842)
• Param : lr: 1e-4, eps: 1e-6, warm_step=10%, epochs: 10, train_batch: 32, eval_batch: 64, max_token_len: 72
• 훈련코드 여기 참조

2.distilation

• 교사 모델 : kpf-sbert-v1.1(max_token_len:128)
• 말뭉치 : news_talk_ko_en_train.tsv (한국어-영어 대화-뉴스 병렬 말뭉치 : 1.38M)
• Param : lr: 5e-5, epochs: 10, train_batch: 32, eval/test_batch: 64, max_token_len: 128(교사모델이 128이므로 맟춰줌)
• 훈련코드 여기 참조

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 256, 'do_lower_case': True}) with Transformer model: AlbertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
)

Citing & Authors

bongsoo