yahyaabd/paraphrase-multilingual-mpnet-base-v2-mnrl

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 statictable-triplets-all 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:
  • statictable-triplets-all

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/paraphrase-multilingual-mpnet-base-v2-mnrl")
# Run inference
sentences = [
    'Seperti apa rincian inflasi di Indonesia, termasuk inflasi inti dan harga diatur, pada 2020?',
    'Inflasi Umum, Inti, Harga Yang Diatur Pemerintah, dan Barang Bergejolak Inflasi Indonesia, 2009-2024',
    'Angka Kematian Ibu/AKI (Maternal Mortality Rate/MMR) Hasil Long Form SP2020 Menurut Provinsi, 2020',
]
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

Information Retrieval

• Dataset: bps-statictable-ir
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9609
cosine_accuracy@5	0.9967
cosine_accuracy@10	1.0
cosine_precision@1	0.9609
cosine_precision@5	0.2339
cosine_precision@10	0.1349
cosine_recall@1	0.7591
cosine_recall@5	0.8053
cosine_recall@10	0.825
cosine_ndcg@1	0.9609
cosine_ndcg@5	0.8586
cosine_ndcg@10	0.854
cosine_mrr@1	0.9609
cosine_mrr@5	0.9769
cosine_mrr@10	0.9773
cosine_map@1	0.9609
cosine_map@5	0.8187
cosine_map@10	0.8083

Training Details

Training Dataset

statictable-triplets-all

• Dataset: statictable-triplets-all at 24979b4
• Size: 967,831 training samples
• Columns: query, pos, and neg
• Approximate statistics based on the first 1000 samples:

  	query	pos	neg
  type	string	string	string
  details	min: 4 tokens mean: 18.38 tokens max: 35 tokens	min: 4 tokens mean: 25.39 tokens max: 50 tokens	min: 4 tokens mean: 25.6 tokens max: 58 tokens

• Samples:

  query	pos	neg
  Neraca arus dana triwulan I tahun 2004 (ringkasan, miliar)	Ringkasan Neraca Arus Dana Triwulan I 2004 (Miliar Rupiah)	Proporsi Penduduk Berumur 10 Tahun ke Atas yang Membaca Surat Kabar/Majalah Selama Seminggu Terakhir menurut Provinsi, Tipe Daerah dan Jenis Kelamin, 2012
  Kumpulan dokumen Rencana Pengurangan Bencana level kabupaten dan kota	Rekap Dokumen RPB Kabupaten/Kota	Penduduk Berumur 15 Tahun Ke Atas yang Bekerja Selama Seminggu yang Lalu Menurut Golongan Umur dan Jumlah Jam Kerja Seluruhnya, 1986-1996
  IHK dan gaji bulanan buruh hotel, di bawah supervisor, 2007=100, tahun 2009	IHK dan Rata-rata Upah per Bulan Buruh Hotel di Bawah Mandor (Supervisor), 2007-2014 (2007=100)	Rata-Rata Bulanan Konsentrasi Partikel Terlarut di Udara Beberapa Kota Menurut Bulan dan Kota (μgr/m3), 2006-2015

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Evaluation Dataset

statictable-triplets-all

• Dataset: statictable-triplets-all at 24979b4
• Size: 967,831 evaluation samples
• Columns: query, pos, and neg
• Approximate statistics based on the first 1000 samples:

  	query	pos	neg
  type	string	string	string
  details	min: 4 tokens mean: 18.39 tokens max: 35 tokens	min: 4 tokens mean: 25.42 tokens max: 58 tokens	min: 4 tokens mean: 25.6 tokens max: 50 tokens

• Samples:

  query	pos	neg
  Bagaimana hubungan IHK dan rata-rata upah buruh hotel (bukan supervisor), acuan 2012, sekitar tahun 2012	IHK dan Rata-rata Upah per Bulan Buruh Hotel di Bawah Mandor (Supervisor), 2012-2014 (2012=100)	Perkembangan Beberapa Agregat Pendapatan dan Pendapatan per Kapita Atas Dasar Harga Berlaku, 2010-2016
  Kegiatan mingguan penduduk 15+ (berdasarkan pendidikan terakhir), 1990	Penduduk Berumur 15 Tahun Ke Atas Menurut Pendidikan Tertinggi yang Ditamatkan dan Jenis Kegiatan Selama Seminggu yang Lalu, 1986-1996	Transaksi Total Atas Dasar Harga Dasar, 2010
  Bandingkan indeks harga konsumen (inflasi) di kota-kota Sumatera vs nasional, Desember 2023	Perbandingan Indeks dan Tingkat Inflasi Desember 2023 Kota-kota di Pulau Sumatera dengan Nasional (2018=100)	Persentase Penduduk Berumur 15 tahun Ke Atas menurut Jenis Kegiatan Seminggu Yang Lalu, 2009-2012

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 1
• warmup_ratio: 0.1
• fp16: True
• load_best_model_at_end: True
• eval_on_start: True
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 1
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: False
• fp16: True
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: True
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• include_for_metrics: []
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• dispatch_batches: None
• split_batches: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: True
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	bps-statictable-ir_cosine_ndcg@10
0	0	-	0.7057	0.5598
0.0348	100	0.2794	0.0554	0.8089
0.0696	200	0.0545	0.0389	0.8189
0.1044	300	0.041	0.0407	0.8194
0.1392	400	0.0381	0.0366	0.8186
0.1740	500	0.0441	0.0283	0.8250
0.2088	600	0.0235	0.0212	0.8405
0.2436	700	0.0216	0.0175	0.8256
0.2784	800	0.0175	0.0119	0.8269
0.3132	900	0.0144	0.0131	0.8086
0.3479	1000	0.008	0.0111	0.8269
0.3827	1100	0.01	0.0106	0.8251
0.4175	1200	0.0238	0.0138	0.8296
0.4523	1300	0.0218	0.0074	0.8360
0.4871	1400	0.0126	0.0077	0.8257
0.5219	1500	0.0082	0.0101	0.8447
0.5567	1600	0.01	0.0057	0.8513
0.5915	1700	0.0057	0.0060	0.8500
0.6263	1800	0.0069	0.0051	0.8522
0.6611	1900	0.0062	0.0053	0.8477
0.6959	2000	0.0056	0.0057	0.8541
0.7307	2100	0.0081	0.0051	0.8492
0.7655	2200	0.0048	0.0049	0.8455
0.8003	2300	0.004	0.0047	0.8493
0.8351	2400	0.0068	0.0041	0.8522
0.8699	2500	0.003	0.0036	0.8530
0.9047	2600	0.0029	0.0035	0.8509
0.9395	2700	0.0031	0.0035	0.8518
0.9743	2800	0.002	0.0034	0.854

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 3.4.1
• Transformers: 4.48.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.3.0
• Datasets: 3.4.0
• 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",
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}