yahyaabd / allstats-semantic-mpnet

SentenceTransformer based on sentence-transformers/paraphrase-multilingual-mpnet-base-v2

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-multilingual-mpnet-base-v2 on the allstats-semantic-dataset-v4 dataset. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: sentence-transformers/paraphrase-multilingual-mpnet-base-v2
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • allstats-semantic-dataset-v4

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("yahyaabd/allstats-semantic-mpnet")
# Run inference
sentences = [
    'Pernikahan usia anak di Indonesia periode 2013-2015',
    'Jumlah penduduk Indonesia 2013-2015',
    'Indeks Tendensi Bisnis dan Indeks Tendensi Konsumen 2013',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Semantic Similarity

• Datasets: allstats-semantic-mpnet-eval and allstats-semantic-mpnet-test
• Evaluated with EmbeddingSimilarityEvaluator

Training Details

Training Dataset

allstats-semantic-dataset-v4

• Dataset: allstats-semantic-dataset-v4 at 06c3cf8
• Size: 88,250 training samples
• Columns: query , doc , and label
• Approximate statistics based on the first 1000 samples:

  	query	doc	label
  type	string	string	float
  details	min: 4 tokens mean: 11.38 tokens max: 46 tokens	min: 4 tokens mean: 14.48 tokens max: 67 tokens	min: 0.0 mean: 0.51 max: 1.0

• Samples:

  query	doc	label
  Industri teh Indonesia tahun 2021	Statistik Transportasi Laut 2014	0.1
  Tahun berapa data pertumbuhan ekonomi Indonesia tersebut?	Nilai Tukar Petani (NTP) November 2023 sebesar 116,73 atau naik 0,82 persen. Harga Gabah Kering Panen di Tingkat Petani turun 1,94 persen dan Harga Beras Premium di Penggilingan turun 0,91 persen.	0.0
  Kemiskinan di Indonesia Maret	2018 Feb Tenaga Kerja	0.1

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Evaluation Dataset

allstats-semantic-dataset-v4

• Dataset: allstats-semantic-dataset-v4 at 06c3cf8
• Size: 18,910 evaluation samples
• Columns: query , doc , and label
• Approximate statistics based on the first 1000 samples:

  	query	doc	label
  type	string	string	float
  details	min: 5 tokens mean: 11.35 tokens max: 33 tokens	min: 4 tokens mean: 14.25 tokens max: 52 tokens	min: 0.0 mean: 0.49 max: 1.0

• Samples:

  query	doc	label
  nAalisis keuangam deas tshun 019	Statistik Migrasi Nusa Tenggara Barat Hasil Survei Penduduk Antar Sensus 2015	0.1
  Data tanaman buah dan sayur Indonesia tahun 2016	Statistik Penduduk Lanjut Usia 2010	0.1
  Pasar beras di Indonesia tahun 2018	Buletin Statistik Perdagangan Luar Negeri Ekspor Menurut Kelompok Komoditi dan Negara, April 2021	0.2

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy : steps
• per_device_train_batch_size : 32
• per_device_eval_batch_size : 32
• num_train_epochs : 8
• warmup_ratio : 0.1
• fp16 : True
• dataloader_num_workers : 4
• load_best_model_at_end : True
• label_smoothing_factor : 0.05
• eval_on_start : True

All Hyperparameters

Training Logs

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.3.1
• Transformers: 4.48.0
• PyTorch: 2.4.1+cu121
• Accelerate: 0.34.2
• Datasets: 3.2.0
• Tokenizers: 0.21.0

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}