cl-nagoya / unsup-simcse-ja-large

unsup-simcse-ja-large

Usage (Sentence-Transformers)

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

pip install -U fugashi[unidic-lite] sentence-transformers

Then you can use the model like this:

from sentence_transformers import SentenceTransformer
sentences = ["こんにちは、世界！", "文埋め込み最高！文埋め込み最高と叫びなさい", "極度乾燥しなさい"]

model = SentenceTransformer("cl-nagoya/unsup-simcse-ja-large")
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("cl-nagoya/unsup-simcse-ja-large")
model = AutoModel.from_pretrained("cl-nagoya/unsup-simcse-ja-large")

# 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)

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (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})
)

Model Summary

• Fine-tuning method: Unsupervised SimCSE
• Base model: cl-tohoku/bert-large-japanese-v2
• Training dataset: Wiki40B
• Pooling strategy: cls (with an extra MLP layer only during training)
• Hidden size: 1024
• Learning rate: 3e-5
• Batch size: 64
• Temperature: 0.05
• Max sequence length: 64
• Number of training examples: 2^20
• Validation interval (steps): 2^6
• Warmup ratio: 0.1
• Dtype: BFloat16

See the GitHub repository for a detailed experimental setup.

Citing & Authors

@misc{
  hayato-tsukagoshi-2023-simple-simcse-ja,
  author = {Hayato Tsukagoshi},
  title = {Japanese Simple-SimCSE},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/hppRC/simple-simcse-ja}}
}