Edit model card

SentenceTransformer based on indobenchmark/indobert-base-p2

This is a sentence-transformers model finetuned from indobenchmark/indobert-base-p2 on the afaji/indonli 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: indobenchmark/indobert-base-p2
  • Maximum Sequence Length: 512 tokens
  • Output Dimensionality: 768 tokens
  • Similarity Function: Cosine Similarity
  • Training Dataset:
  • Language: id

Model Sources

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': 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("cassador/indobert-base-p2-nli-v2")
# Run inference
sentences = [
    'Set album musik pengiring seri film Harry Potter akan dirilis dalam versi baru.',
    'Seri film Harry Potter memiliki set album musik pengiring.',
    'Laga dan kolosal adalah genre film.',
]
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

Metric Value
pearson_cosine 0.3021
spearman_cosine 0.303
pearson_manhattan 0.2768
spearman_manhattan 0.2726
pearson_euclidean 0.3072
spearman_euclidean 0.3045
pearson_dot 0.3039
spearman_dot 0.3047
pearson_max 0.3072
spearman_max 0.3047

Semantic Similarity

Metric Value
pearson_cosine 0.1038
spearman_cosine 0.0969
pearson_manhattan 0.0749
spearman_manhattan 0.0782
pearson_euclidean 0.0942
spearman_euclidean 0.099
pearson_dot 0.107
spearman_dot 0.0998
pearson_max 0.107
spearman_max 0.0998

Training Details

Training Dataset

afaji/indonli

  • Dataset: afaji/indonli
  • Size: 10,000 training samples
  • Columns: premise, hypothesis, and label
  • Approximate statistics based on the first 1000 samples:
    premise hypothesis label
    type string string int
    details
    • min: 12 tokens
    • mean: 29.73 tokens
    • max: 179 tokens
    • min: 6 tokens
    • mean: 11.93 tokens
    • max: 35 tokens
    • 0: ~68.60%
    • 1: ~31.40%
  • Samples:
    premise hypothesis label
    Presiden Joko Widodo (Jokowi) menyampaikan prediksi bahwa wabah virus Corona (COVID-19) di Indonesia akan selesai akhir tahun ini. Prediksi akhir wabah tidak disampaikan Jokowi. 0
    Meski biasanya hanya digunakan di fasilitas kesehatan, saat ini masker dan sarung tangan sekali pakai banyak dipakai di tingkat rumah tangga. Masker sekali pakai banyak dipakai di tingkat rumah tangga. 1
    Data dari Nielsen Music mencatat, "Joanne" telah terjual 201 ribu kopi di akhir minggu ini, seperti dilansir aceshowbiz.com. Nielsen Music mencatat pada akhir minggu ini. 0
  • Loss: SoftmaxLoss

Evaluation Dataset

afaji/indonli

  • Dataset: afaji/indonli
  • Size: 2,000 evaluation samples
  • Columns: premise, hypothesis, and label
  • Approximate statistics based on the first 1000 samples:
    premise hypothesis label
    type string string int
    details
    • min: 9 tokens
    • mean: 28.09 tokens
    • max: 179 tokens
    • min: 6 tokens
    • mean: 12.01 tokens
    • max: 24 tokens
    • 0: ~63.00%
    • 1: ~37.00%
  • Samples:
    premise hypothesis label
    Manuskrip tersebut berisi tiga catatan yang menceritakan bagaimana peristiwa jatuhnya meteorit serta laporan kematian akibat kejadian tersebut seperti dilansir dari Science Alert, Sabtu (25/4/2020). Manuskrip tersebut tidak mencatat laporan kematian. 0
    Dilansir dari Business Insider, menurut observasi dari Mauna Loa Observatory di Hawaii pada karbon dioksida (CO2) di level mencapai 410 ppm tidak langsung memberikan efek pada pernapasan, karena tubuh manusia juga masih membutuhkan CO2 dalam kadar tertentu. Tidak ada observasi yang pernah dilansir oleh Business Insider. 0
    Perekonomian Jakarta terutama ditunjang oleh sektor perdagangan, jasa, properti, industri kreatif, dan keuangan. Sektor jasa memberi pengaruh lebih besar daripada industri kreatif dalam perekonomian Jakarta. 0
  • Loss: SoftmaxLoss

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: epoch
  • learning_rate: 1e-05
  • num_train_epochs: 10
  • warmup_ratio: 0.001
  • fp16: True

All Hyperparameters

Click to expand
  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: epoch
  • prediction_loss_only: True
  • per_device_train_batch_size: 8
  • per_device_eval_batch_size: 8
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 1
  • eval_accumulation_steps: None
  • learning_rate: 1e-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: 10
  • max_steps: -1
  • lr_scheduler_type: linear
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.001
  • 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: False
  • 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: False
  • hub_always_push: False
  • gradient_checkpointing: False
  • gradient_checkpointing_kwargs: None
  • include_inputs_for_metrics: False
  • 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
  • batch_sampler: batch_sampler
  • multi_dataset_batch_sampler: proportional

Training Logs

Click to expand
Epoch Step Training Loss loss sts-dev_spearman_cosine sts-test_spearman_cosine
0 0 - - 0.1928 -
0.04 100 1.1407 - - -
0.08 200 0.7456 - - -
0.12 300 0.6991 - - -
0.16 400 0.6653 - - -
0.2 500 0.6317 - - -
0.24 600 0.5975 - - -
0.28 700 0.5955 - - -
0.32 800 0.6168 - - -
0.36 900 0.5851 - - -
0.4 1000 0.591 - - -
0.44 1100 0.6063 - - -
0.48 1200 0.6122 - - -
0.52 1300 0.5881 - - -
0.56 1400 0.59 - - -
0.6 1500 0.5715 - - -
0.64 1600 0.5725 - - -
0.68 1700 0.5771 - - -
0.72 1800 0.5935 - - -
0.76 1900 0.584 - - -
0.8 2000 0.5829 - - -
0.84 2100 0.5507 - - -
0.88 2200 0.5447 - - -
0.92 2300 0.6059 - - -
0.96 2400 0.5389 - - -
1.0 2500 0.639 0.5432 0.4007 -
1.04 2600 0.463 - - -
1.08 2700 0.4936 - - -
1.12 2800 0.4966 - - -
1.16 2900 0.4588 - - -
1.2 3000 0.5148 - - -
1.24 3100 0.5043 - - -
1.28 3200 0.5048 - - -
1.32 3300 0.4803 - - -
1.3600 3400 0.465 - - -
1.4 3500 0.5133 - - -
1.44 3600 0.5505 - - -
1.48 3700 0.4498 - - -
1.52 3800 0.5418 - - -
1.56 3900 0.5268 - - -
1.6 4000 0.4546 - - -
1.6400 4100 0.5279 - - -
1.6800 4200 0.5309 - - -
1.72 4300 0.487 - - -
1.76 4400 0.5371 - - -
1.8 4500 0.5097 - - -
1.8400 4600 0.5242 - - -
1.88 4700 0.4583 - - -
1.92 4800 0.4923 - - -
1.96 4900 0.5028 - - -
2.0 5000 0.5139 0.6274 0.4335 -
2.04 5100 0.322 - - -
2.08 5200 0.389 - - -
2.12 5300 0.3633 - - -
2.16 5400 0.3868 - - -
2.2 5500 0.3798 - - -
2.24 5600 0.4385 - - -
2.2800 5700 0.3965 - - -
2.32 5800 0.3895 - - -
2.36 5900 0.4484 - - -
2.4 6000 0.3452 - - -
2.44 6100 0.3905 - - -
2.48 6200 0.376 - - -
2.52 6300 0.4986 - - -
2.56 6400 0.3732 - - -
2.6 6500 0.3632 - - -
2.64 6600 0.3915 - - -
2.68 6700 0.4394 - - -
2.7200 6800 0.3852 - - -
2.76 6900 0.3984 - - -
2.8 7000 0.426 - - -
2.84 7100 0.3274 - - -
2.88 7200 0.4673 - - -
2.92 7300 0.4599 - - -
2.96 7400 0.4304 - - -
3.0 7500 0.4151 0.8967 0.4007 -
3.04 7600 0.2345 - - -
3.08 7700 0.1807 - - -
3.12 7800 0.2984 - - -
3.16 7900 0.2357 - - -
3.2 8000 0.4506 - - -
3.24 8100 0.2178 - - -
3.2800 8200 0.2654 - - -
3.32 8300 0.2863 - - -
3.36 8400 0.2626 - - -
3.4 8500 0.3281 - - -
3.44 8600 0.2555 - - -
3.48 8700 0.4245 - - -
3.52 8800 0.2368 - - -
3.56 8900 0.3288 - - -
3.6 9000 0.3417 - - -
3.64 9100 0.3249 - - -
3.68 9200 0.3378 - - -
3.7200 9300 0.233 - - -
3.76 9400 0.3215 - - -
3.8 9500 0.251 - - -
3.84 9600 0.3138 - - -
3.88 9700 0.3081 - - -
3.92 9800 0.3875 - - -
3.96 9900 0.3231 - - -
4.0 10000 0.2119 1.4983 0.4129 -
4.04 10100 0.1323 - - -
4.08 10200 0.2222 - - -
4.12 10300 0.2005 - - -
4.16 10400 0.127 - - -
4.2 10500 0.1052 - - -
4.24 10600 0.1657 - - -
4.28 10700 0.2305 - - -
4.32 10800 0.1048 - - -
4.36 10900 0.2081 - - -
4.4 11000 0.201 - - -
4.44 11100 0.1515 - - -
4.48 11200 0.2112 - - -
4.52 11300 0.1936 - - -
4.5600 11400 0.1578 - - -
4.6 11500 0.2551 - - -
4.64 11600 0.2888 - - -
4.68 11700 0.128 - - -
4.72 11800 0.2172 - - -
4.76 11900 0.114 - - -
4.8 12000 0.2135 - - -
4.84 12100 0.2421 - - -
4.88 12200 0.2392 - - -
4.92 12300 0.1478 - - -
4.96 12400 0.1901 - - -
5.0 12500 0.2219 1.9582 0.3469 -
5.04 12600 0.1586 - - -
5.08 12700 0.1587 - - -
5.12 12800 0.0663 - - -
5.16 12900 0.0703 - - -
5.2 13000 0.0783 - - -
5.24 13100 0.1143 - - -
5.28 13200 0.1155 - - -
5.32 13300 0.0661 - - -
5.36 13400 0.0935 - - -
5.4 13500 0.1344 - - -
5.44 13600 0.1031 - - -
5.48 13700 0.1294 - - -
5.52 13800 0.103 - - -
5.5600 13900 0.0739 - - -
5.6 14000 0.1477 - - -
5.64 14100 0.1171 - - -
5.68 14200 0.1504 - - -
5.72 14300 0.1122 - - -
5.76 14400 0.1279 - - -
5.8 14500 0.0813 - - -
5.84 14600 0.1372 - - -
5.88 14700 0.1615 - - -
5.92 14800 0.1944 - - -
5.96 14900 0.0436 - - -
6.0 15000 0.1195 2.2220 0.3559 -
0.08 100 0.0844 - - -
0.16 200 0.1357 - - -
0.24 300 0.1382 - - -
0.32 400 0.2091 - - -
0.4 500 0.2351 - - -
0.48 600 0.2976 - - -
0.56 700 0.3408 - - -
0.64 800 0.2656 - - -
0.72 900 0.3183 - - -
0.8 1000 0.2513 - - -
0.88 1100 0.2293 - - -
0.96 1200 0.3241 - - -
1.0 1250 - 1.1813 0.3495 -
0.3195 100 0.6132 - - -
0.6390 200 0.1554 - - -
0.9585 300 0.1366 - - -
1.0 313 - 1.2867 0.3839 -
0.08 100 0.2713 - - -
0.16 200 0.1273 - - -
0.24 300 0.0883 - - -
0.32 400 0.0749 - - -
0.08 100 0.0653 - - -
0.16 200 0.0311 - - -
0.24 300 0.0368 - - -
0.32 400 0.0259 - - -
0.4 500 0.059 - - -
0.48 600 0.046 - - -
0.56 700 0.1266 - - -
0.64 800 0.0661 - - -
0.72 900 0.0676 - - -
0.8 1000 0.0759 - - -
0.88 1100 0.0527 - - -
0.96 1200 0.1038 - - -
1.0 1250 - 2.2411 0.3892 -
1.04 1300 0.0456 - - -
1.12 1400 0.1363 - - -
1.2 1500 0.1398 - - -
1.28 1600 0.1237 - - -
1.3600 1700 0.123 - - -
1.44 1800 0.1893 - - -
1.52 1900 0.1192 - - -
1.6 2000 0.1347 - - -
1.6800 2100 0.0937 - - -
1.76 2200 0.1506 - - -
1.8400 2300 0.1366 - - -
1.92 2400 0.1194 - - -
2.0 2500 0.1485 2.1340 0.3245 -
2.08 2600 0.0485 - - -
2.16 2700 0.0579 - - -
2.24 2800 0.0932 - - -
2.32 2900 0.0743 - - -
2.4 3000 0.0783 - - -
2.48 3100 0.0918 - - -
2.56 3200 0.0973 - - -
2.64 3300 0.0623 - - -
2.7200 3400 0.1284 - - -
2.8 3500 0.1247 - - -
2.88 3600 0.0648 - - -
2.96 3700 0.0921 - - -
3.0 3750 - 2.4354 0.2824 -
3.04 3800 0.04 - - -
3.12 3900 0.0417 - - -
3.2 4000 0.0414 - - -
3.2800 4100 0.0485 - - -
3.36 4200 0.0255 - - -
3.44 4300 0.0688 - - -
3.52 4400 0.0574 - - -
3.6 4500 0.0766 - - -
3.68 4600 0.0481 - - -
3.76 4700 0.06 - - -
3.84 4800 0.0528 - - -
3.92 4900 0.0426 - - -
4.0 5000 0.092 2.5427 0.3284 -
4.08 5100 0.0349 - - -
4.16 5200 0.0107 - - -
4.24 5300 0.0608 - - -
4.32 5400 0.0473 - - -
4.4 5500 0.0452 - - -
4.48 5600 0.0316 - - -
4.5600 5700 0.0096 - - -
4.64 5800 0.0511 - - -
4.72 5900 0.0207 - - -
4.8 6000 0.0061 - - -
4.88 6100 0.0381 - - -
4.96 6200 0.0378 - - -
5.0 6250 - 2.6061 0.3061 -
5.04 6300 0.0326 - - -
5.12 6400 0.0349 - - -
5.2 6500 0.0128 - - -
5.28 6600 0.0185 - - -
5.36 6700 0.0145 - - -
5.44 6800 0.0521 - - -
5.52 6900 0.0427 - - -
5.6 7000 0.0215 - - -
5.68 7100 0.0195 - - -
5.76 7200 0.0426 - - -
5.84 7300 0.057 - - -
5.92 7400 0.0106 - - -
6.0 7500 0.0284 2.8348 0.3291 -
6.08 7600 0.0286 - - -
6.16 7700 0.018 - - -
6.24 7800 0.0224 - - -
6.32 7900 0.0102 - - -
6.4 8000 0.0287 - - -
6.48 8100 0.0078 - - -
6.5600 8200 0.0237 - - -
6.64 8300 0.0148 - - -
6.72 8400 0.0271 - - -
6.8 8500 0.015 - - -
6.88 8600 0.0278 - - -
6.96 8700 0.0237 - - -
7.0 8750 - 2.8785 0.3188 -
7.04 8800 0.0203 - - -
7.12 8900 0.0089 - - -
7.2 9000 0.0121 - - -
7.28 9100 0.0185 - - -
7.36 9200 0.0127 - - -
7.44 9300 0.017 - - -
7.52 9400 0.0117 - - -
7.6 9500 0.006 - - -
7.68 9600 0.0061 - - -
7.76 9700 0.0141 - - -
7.84 9800 0.0091 - - -
7.92 9900 0.0164 - - -
8.0 10000 0.0244 2.8054 0.3040 -
8.08 10100 0.0001 - - -
8.16 10200 0.0187 - - -
8.24 10300 0.0098 - - -
8.32 10400 0.0114 - - -
8.4 10500 0.004 - - -
8.48 10600 0.0017 - - -
8.56 10700 0.0018 - - -
8.64 10800 0.009 - - -
8.72 10900 0.0047 - - -
8.8 11000 0.0014 - - -
8.88 11100 0.0049 - - -
8.96 11200 0.006 - - -
9.0 11250 - 2.9460 0.2967 -
9.04 11300 0.0057 - - -
9.12 11400 0.0051 - - -
9.2 11500 0.0067 - - -
9.28 11600 0.0009 - - -
9.36 11700 0.0046 - - -
9.44 11800 0.0138 - - -
9.52 11900 0.0067 - - -
9.6 12000 0.0043 - - -
9.68 12100 0.001 - - -
9.76 12200 0.0004 - - -
9.84 12300 0.0044 - - -
9.92 12400 0.003 - - -
10.0 12500 0.0055 2.9714 0.3030 0.0969

Framework Versions

  • Python: 3.10.12
  • Sentence Transformers: 3.0.1
  • Transformers: 4.41.2
  • PyTorch: 2.3.0+cu121
  • Accelerate: 0.31.0
  • Datasets: 2.20.0
  • Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers and SoftmaxLoss

@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",
}
Downloads last month
2
Safetensors
Model size
124M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for cassador/indobert-base-p2-nli-v2

Finetuned
(35)
this model

Dataset used to train cassador/indobert-base-p2-nli-v2

Evaluation results