README.md

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:139719
  - loss:CategoricalContrastiveLoss
widget:
  - source_sentence: 科目：コンクリート。名称：底盤ｺﾝｸﾘｰﾄ打設手間。
    sentences:
      - >-
        科目：コンクリート。名称：基礎部ﾏｽｺﾝｸﾘｰﾄ。摘要：FC36 S15粗骨材20 高性能AE減水剤高炉ｾﾒﾝﾄB種。備考：代価表   
        0103。
      - >-
        科目：コンクリート。名称：基礎部ｺﾝｸﾘｰﾄ。摘要：FC36N/mm2
        ｽﾗﾝﾌﾟ18高性能AE減水剤ﾏｽｺﾝｸﾘｰﾄ中庸熱ﾎﾟﾙﾄﾗﾝﾄﾞｾﾒﾝﾄ。備考：代価表   
        0031S-01厚さ1000mm以上の耐圧ｽﾗﾌﾞ､梁幅800mm以上の基礎梁。
      - 科目：コンクリート。名称：ｺﾝｸﾘｰﾄ打設手間。
  - source_sentence: 科目：コンクリート。名称：立上り壁ｺﾝｸﾘｰﾄ。
    sentences:
      - 科目：コンクリート。名称：機械基礎ｺﾝｸﾘｰﾄ。摘要：FC21N/mm2 ｽﾗﾝﾌﾟ15。備考：代価表    0045。
      - 科目：タイル。名称：ﾄﾞﾗｲｴﾘｱ床ﾀｲﾙ。
      - 科目：コンクリート。名称：ｵｲﾙﾀﾝｸ基礎ｺﾝｸﾘｰﾄ。摘要：FC24 S18粗骨材20 高性能AE減水剤。備考：代価表    0108。
  - source_sentence: 科目：タイル。名称：昇降口床ﾀｲﾙ。
    sentences:
      - 科目：タイル。名称：ｱﾌﾟﾛｰﾁﾃﾗｽ立上り床ﾀｲﾙA。
      - 科目：タイル。名称：昇降口床ﾀｲﾙ張り。
      - 科目：タイル。名称：ﾋﾟﾛﾃｨ床床ﾀｲﾙA。
  - source_sentence: 科目：コンクリート。名称：免震BPL下部充填ｺﾝｸﾘｰﾄ打設手間。
    sentences:
      - 科目：コンクリート。名称：充填ｺﾝｸﾘｰﾄ（EXP_J内）。
      - 科目：タイル。名称：地流し壁小口ﾀｲﾙ。
      - 科目：タイル。名称：地流し床ﾀｲﾙ。
  - source_sentence: 科目：コンクリート。名称：基礎部ﾏｽｺﾝｸﾘｰﾄ。
    sentences:
      - 科目：コンクリート。名称：ﾎﾟﾝﾌﾟ圧送。
      - 科目：コンクリート。名称：基礎部ｺﾝｸﾘｰﾄ。摘要：JIS A5308   呼び強度36  S15粗骨材20。備考：刊-CON K3615。
      - 科目：コンクリート。名称：基礎部普通ｺﾝｸﾘｰﾄ。摘要：FC30 S15AE減水剤。備考：コンクリー 1。
pipeline_tag: sentence-similarity
library_name: sentence-transformers

SentenceTransformer

This is a sentence-transformers model trained. 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
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity

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': 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, '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("Detomo/cl-nagoya-sup-simcse-ja-nss-v_1_0_7_1")
# Run inference
sentences = [
    '科目：コンクリート。名称：基礎部ﾏｽｺﾝｸﾘｰﾄ。',
    '科目：コンクリート。名称：基礎部普通ｺﾝｸﾘｰﾄ。摘要：FC30 S15AE減水剤。備考：コンクリー 1。',
    '科目：コンクリート。名称：ﾎﾟﾝﾌﾟ圧送。',
]
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]

Training Details

Training Dataset

Unnamed Dataset

• Size: 139,719 training samples
• Columns: sentence1, sentence2, and label
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	int
  details	min: 11 tokens mean: 14.03 tokens max: 19 tokens	min: 11 tokens mean: 22.75 tokens max: 72 tokens	0: ~12.60% 1: ~8.60% 2: ~78.80%

• Samples:

  sentence1	sentence2	label
  科目：コンクリート。名称：ｺﾝｸﾘｰﾄﾎﾟﾝﾌﾟ圧送。	科目：コンクリート。名称：ﾎﾟﾝﾌﾟ圧送。	1
  科目：コンクリート。名称：ｺﾝｸﾘｰﾄﾎﾟﾝﾌﾟ圧送。	科目：コンクリート。名称：充填ｺﾝｸﾘｰﾄ（EXP_J内）。摘要：Fc18N/mm2 S18。備考：刊-ｺﾝｸﾘｰﾄ 1818物P100×100%。	0
  科目：コンクリート。名称：ｺﾝｸﾘｰﾄﾎﾟﾝﾌﾟ圧送。	科目：コンクリート。名称：EXP_J充填ｺﾝｸﾘｰﾄ。	0

• Loss: sentence_transformer_lib.categorical_constrastive_loss.CategoricalContrastiveLoss

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 512
• per_device_eval_batch_size: 512
• learning_rate: 1e-05
• weight_decay: 0.01
• num_train_epochs: 20
• warmup_ratio: 0.2
• fp16: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 512
• per_device_eval_batch_size: 512
• 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: 1e-05
• weight_decay: 0.01
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 20
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.2
• 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}
• tp_size: 0
• 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
• 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: False
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss
0.1832	50	0.6905
0.3663	100	0.2528
0.5495	150	0.1824
0.7326	200	0.1544
0.9158	250	0.14
1.0989	300	0.1272
1.2821	350	0.1135
1.4652	400	0.1001
1.6484	450	0.0987
1.8315	500	0.0887
2.0147	550	0.0804
2.1978	600	0.074
2.3810	650	0.0713
2.5641	700	0.0666
2.7473	750	0.06
2.9304	800	0.0601
3.1136	850	0.0494
3.2967	900	0.0472
3.4799	950	0.046
3.6630	1000	0.0441
3.8462	1050	0.0416
4.0293	1100	0.0373
4.2125	1150	0.034
4.3956	1200	0.0308
4.5788	1250	0.0308
4.7619	1300	0.0311
4.9451	1350	0.0273
5.1282	1400	0.0225
5.3114	1450	0.0231
5.4945	1500	0.0218
5.6777	1550	0.0209
5.8608	1600	0.0193
6.0440	1650	0.0182
6.2271	1700	0.0161
6.4103	1750	0.0161
6.5934	1800	0.0162
6.7766	1850	0.0146
6.9597	1900	0.0146
7.1429	1950	0.0126
7.3260	2000	0.0118
7.5092	2050	0.012
7.6923	2100	0.0118
7.8755	2150	0.0116
8.0586	2200	0.0121
8.2418	2250	0.0098
8.4249	2300	0.0099
8.6081	2350	0.0094
8.7912	2400	0.0089
8.9744	2450	0.009
9.1575	2500	0.0079
9.3407	2550	0.0082
9.5238	2600	0.0077
9.7070	2650	0.0074
9.8901	2700	0.008
10.0733	2750	0.0074
10.2564	2800	0.0065
10.4396	2850	0.0069
10.6227	2900	0.0067
10.8059	2950	0.0063
10.9890	3000	0.0064
11.1722	3050	0.0057
11.3553	3100	0.0058
11.5385	3150	0.0055
11.7216	3200	0.005
11.9048	3250	0.0055
12.0879	3300	0.0049
12.2711	3350	0.0041
12.4542	3400	0.0045
12.6374	3450	0.0045
12.8205	3500	0.0052
13.0037	3550	0.0054
13.1868	3600	0.005
13.3700	3650	0.0041
13.5531	3700	0.0039
13.7363	3750	0.004
13.9194	3800	0.0043
14.1026	3850	0.0037
14.2857	3900	0.0036
14.4689	3950	0.0038
14.6520	4000	0.0037
14.8352	4050	0.0042
15.0183	4100	0.004
15.2015	4150	0.0036
15.3846	4200	0.0036
15.5678	4250	0.0032
15.7509	4300	0.0032
15.9341	4350	0.0028
16.1172	4400	0.0032
16.3004	4450	0.0027
16.4835	4500	0.0034
16.6667	4550	0.0035
16.8498	4600	0.0032
17.0330	4650	0.0035
17.2161	4700	0.0031
17.3993	4750	0.003
17.5824	4800	0.003
17.7656	4850	0.0029
17.9487	4900	0.0029
18.1319	4950	0.0022
18.3150	5000	0.0034
18.4982	5050	0.0028
18.6813	5100	0.0026
18.8645	5150	0.0028
19.0476	5200	0.0025
19.2308	5250	0.0027
19.4139	5300	0.0029
19.5971	5350	0.0026
19.7802	5400	0.0027
19.9634	5450	0.0029

Framework Versions

• Python: 3.11.12
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.6.0
• Datasets: 2.14.4
• Tokenizers: 0.21.1

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",
}