Albert Small Kor Sbert V1

🚀 albert-small-kor-sbert-v1

This is a sentence-transformers model that maps sentences and paragraphs to a 768-dimensional dense vector space. It can be used for tasks such as clustering or semantic search. This model is developed based on the albert-small-kor-v1 model using the SentenceBERT approach.

🚀 Quick Start

📦 Installation

You can install the necessary library with the following command:

pip install -U sentence-transformers

💻 Usage Examples

Basic Usage

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

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

Advanced Usage

Without sentence-transformers, you can use the model as follows. First, pass your input through the transformer model, then apply the appropriate 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')
model = AutoModel.from_pretrained('bongsoo/albert-small-kor-sbert-v1')

# 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)

📚 Documentation

Evaluation Results

• For performance measurement, the following Korean (kor) and English (en) evaluation corpora were used:
  • Korean: korsts (1,379 sentence pairs) and klue-sts (519 sentence pairs)
  • English: stsb_multi_mt (1,376 sentence pairs) and glue:stsb (1,500 sentence pairs)
• The performance metric is cosin.spearman.
• Refer to the evaluation code here.

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

Training

The albert-small-kor-v1 model was trained using the sts(10)-distil(10)-nli(3)-sts(10) approach.

Common Parameters

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

1. STS

• Corpus: korsts (5,749) + kluestsV1.1 (11,668) + stsb_multi_mt (5,749) + mteb/sickr-sts (9,927) + glue stsb (5,749) (Total: 38,842)
• Parameters: lr: 1e-4, eps: 1e-6, warm_step=10%, epochs: 10, train_batch: 32, eval_batch: 64, max_token_len: 72
• Refer to the training code here.

2. Distillation

• Teacher model: paraphrase-multilingual-mpnet-base-v2 (max_token_len: 128)
• Corpus: news_talk_en_ko_train.tsv (English-Korean dialogue-news parallel corpus: 1.38M)
• Parameters: lr: 5e-5, eps: 1e-8, epochs: 10, train_batch: 32, eval/test_batch: 64, max_token_len: 128 (to match the teacher model)
• Refer to the training code here.

3. NLI

• Corpus: Training (967,852): kornli (550,152), kluenli (24,998), glue-mnli (392,702); Evaluation (3,519): korsts (1,500), kluests (519), gluests (1,500)
• Hyperparameters: lr: 3e-5, eps: 1e-8, warm_step=10%, epochs: 3, train/eval_batch: 64, max_token_len: 128
• Refer to the training code here.

🔧 Technical Details

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})
)

📄 License

No license information provided.

Citing & Authors

bongsoo