--- language: [] library_name: sentence-transformers tags: - sentence-transformers - sentence-similarity - feature-extraction - dataset_size:100K - **Maximum Sequence Length:** 8192 tokens - **Output Dimensionality:** 768 tokens - **Similarity Function:** Cosine Similarity ### Model Sources - **Documentation:** [Sentence Transformers Documentation](https://sbert.net) - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers) - **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers) ### Full Model Architecture ``` SentenceTransformer( (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: NomicBertModel (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: ```bash pip install -U sentence-transformers ``` Then you can load this model and run inference. ```python from sentence_transformers import SentenceTransformer # Download from the 🤗 Hub model = SentenceTransformer("sentence_transformers_model_id") # Run inference sentences = [ 'search_query: お布団バッグ', 'search_query: 足なしソファー', 'search_query: all color handbag', ] 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 #### Triplet * Dataset: `triplet-esci` * Evaluated with [TripletEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.TripletEvaluator) | Metric | Value | |:--------------------|:----------| | **cosine_accuracy** | **0.787** | | dot_accuracy | 0.22 | | manhattan_accuracy | 0.762 | | euclidean_accuracy | 0.768 | | max_accuracy | 0.787 | ## Training Details ### Training Dataset #### Unnamed Dataset * Size: 100,000 training samples * Columns: anchor, positive, and negative * Approximate statistics based on the first 1000 samples: | | anchor | positive | negative | |:--------|:----------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------| | type | string | string | string | | details | | | | * Samples: | anchor | positive | negative | |:------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------| | search_query: blー5c | search_document: [EnergyPower] TECSUN PL-368 電池2個セット SSB・同期検波・長波 [交換用バッテリーBL-5C付] デジタルDSPポケット短波ラジオ 超小型 長・中波用外付アンテナ 10キー ポータブルBCL受信機 FMステレオ/LW/MW/SW ワールドバンドレシーバー 850局プリセットメモリー シグナルメーター USB充電 スリープタイマー アラー, TECSUN, PL-368 電池＋セット [ブラック] | search_document: RADIWOWで作る SIHUADON R108 ポータブル BCL短波ラジオAM FM LW SW 航空無線 DSPレシーバー LCD 良好屋内および屋外アクティビティの両親への贈り物, RADIWOW, グレー | | search_query: かわいいロングtシャツ | search_document: レディース ロンt 半袖 tシャツ オーバーサイズ コットン スリット 大きいサイズ 白 シャツ ビッグシルエット ワンピース シャツワンピ ロングｔシャツ おおきいサイズ 夏 ピンク カジュアル カップ付き カーディガン キラキラ キャミソール キャミ サテン シンプル シニア シフォン シースルー シ, Sleeping Sheep(スリーピング シープ), ホワイト | search_document: Perkisboby スポーツウェア レディース ヨガウェア 4点セット 上下セット 5点セットウェア フィットネス 2点セット ジャージ スポーツブラ パンツ パーカー 半袖 ハーフパンツ, Perkisboby, 2点セット-グレー | | search_query: iphone xr otterbox symmetry case | search_document: Symmetry Clear Series Case for iPhone XR (ONLY) Symmetry Case for iPhone XR Symmetry Case - Clear, VTSOU, Clear | search_document: OtterBox Symmetry Series Case for Apple iPhone XS Max - Tonic Violet / Purple, OtterBox, Tonic Violet / Purple | * Loss: [CachedMultipleNegativesRankingLoss](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cachedmultiplenegativesrankingloss) with these parameters: ```json { "scale": 20.0, "similarity_fct": "cos_sim" } ``` ### Evaluation Dataset #### Unnamed Dataset * Size: 1,000 evaluation samples * Columns: anchor, positive, and negative * Approximate statistics based on the first 1000 samples: | | anchor | positive | negative | |:--------|:----------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------| | type | string | string | string | | details | | | | * Samples: | anchor | positive | negative | |:---------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------| | search_query: snack vending machine | search_document: Red All Metal Triple Compartment Commercial Vending Machine for 1 inch Gumballs, 1 inch Toy Capsules, Bouncy Balls, Candy, Nuts with Stand by American Gumball Company, American Gumball Company, CANDY RED | search_document: Vending Machine Halloween Costume - Funny Snack Food Adult Men & Women Outfits, Hauntlook, Multicolored | | search_query: slim credit card holder without id window | search_document: Banuce Top Grain Leather Card Holder for Women Men Unisex ID Credit Card Case Slim Card Wallet Black, Banuce, 1 ID + 5 Card Slots: Black | search_document: Mens Wallet RFID Genuine Leather Bifold Wallets For Men, ID Window 16 Card Holders Gift Box, Swallowmall, Black Stripe | | search_query: gucci belts for women | search_document: Gucci Women's Gg0027o 50Mm Optical Glasses, Gucci, Havana | search_document: Gucci G-Gucci Gold PVD Women's Watch(Model:YA125511), Gucci, PVD/Brown | * Loss: [CachedMultipleNegativesRankingLoss](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cachedmultiplenegativesrankingloss) with these parameters: ```json { "scale": 20.0, "similarity_fct": "cos_sim" } ``` ### Training Hyperparameters #### Non-Default Hyperparameters - `per_device_train_batch_size`: 4 - `per_device_eval_batch_size`: 4 - `gradient_accumulation_steps`: 2 - `learning_rate`: 1e-06 - `lr_scheduler_type`: cosine - `warmup_ratio`: 0.1 - `dataloader_drop_last`: True - `dataloader_num_workers`: 4 - `dataloader_prefetch_factor`: 2 - `load_best_model_at_end`: True - `batch_sampler`: no_duplicates #### All Hyperparameters
Click to expand - `overwrite_output_dir`: False - `do_predict`: False - `prediction_loss_only`: True - `per_device_train_batch_size`: 4 - `per_device_eval_batch_size`: 4 - `per_gpu_train_batch_size`: None - `per_gpu_eval_batch_size`: None - `gradient_accumulation_steps`: 2 - `eval_accumulation_steps`: None - `learning_rate`: 1e-06 - `weight_decay`: 0.0 - `adam_beta1`: 0.9 - `adam_beta2`: 0.999 - `adam_epsilon`: 1e-08 - `max_grad_norm`: 1.0 - `num_train_epochs`: 3 - `max_steps`: -1 - `lr_scheduler_type`: cosine - `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 - `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`: False - `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`: True - `dataloader_num_workers`: 4 - `dataloader_prefetch_factor`: 2 - `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} - `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 - `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 - `batch_sampler`: no_duplicates - `multi_dataset_batch_sampler`: proportional
### Training Logs
Click to expand | Epoch | Step | Training Loss | loss | triplet-esci_cosine_accuracy | |:------:|:-----:|:-------------:|:------:|:----------------------------:| | 0.008 | 100 | 0.7191 | - | - | | 0.016 | 200 | 0.6917 | - | - | | 0.024 | 300 | 0.7129 | - | - | | 0.032 | 400 | 0.6826 | - | - | | 0.04 | 500 | 0.7317 | - | - | | 0.048 | 600 | 0.7237 | - | - | | 0.056 | 700 | 0.6904 | - | - | | 0.064 | 800 | 0.6815 | - | - | | 0.072 | 900 | 0.6428 | - | - | | 0.08 | 1000 | 0.6561 | 0.6741 | 0.74 | | 0.088 | 1100 | 0.6097 | - | - | | 0.096 | 1200 | 0.6426 | - | - | | 0.104 | 1300 | 0.618 | - | - | | 0.112 | 1400 | 0.6346 | - | - | | 0.12 | 1500 | 0.611 | - | - | | 0.128 | 1600 | 0.6092 | - | - | | 0.136 | 1700 | 0.6512 | - | - | | 0.144 | 1800 | 0.646 | - | - | | 0.152 | 1900 | 0.6584 | - | - | | 0.16 | 2000 | 0.6403 | 0.6411 | 0.747 | | 0.168 | 2100 | 0.5882 | - | - | | 0.176 | 2200 | 0.6361 | - | - | | 0.184 | 2300 | 0.5641 | - | - | | 0.192 | 2400 | 0.5734 | - | - | | 0.2 | 2500 | 0.6156 | - | - | | 0.208 | 2600 | 0.6252 | - | - | | 0.216 | 2700 | 0.634 | - | - | | 0.224 | 2800 | 0.5743 | - | - | | 0.232 | 2900 | 0.5222 | - | - | | 0.24 | 3000 | 0.5604 | 0.6180 | 0.765 | | 0.248 | 3100 | 0.5864 | - | - | | 0.256 | 3200 | 0.5541 | - | - | | 0.264 | 3300 | 0.5661 | - | - | | 0.272 | 3400 | 0.5493 | - | - | | 0.28 | 3500 | 0.556 | - | - | | 0.288 | 3600 | 0.56 | - | - | | 0.296 | 3700 | 0.5552 | - | - | | 0.304 | 3800 | 0.5833 | - | - | | 0.312 | 3900 | 0.5578 | - | - | | 0.32 | 4000 | 0.5495 | 0.6009 | 0.769 | | 0.328 | 4100 | 0.5245 | - | - | | 0.336 | 4200 | 0.477 | - | - | | 0.344 | 4300 | 0.5536 | - | - | | 0.352 | 4400 | 0.5493 | - | - | | 0.36 | 4500 | 0.532 | - | - | | 0.368 | 4600 | 0.5341 | - | - | | 0.376 | 4700 | 0.528 | - | - | | 0.384 | 4800 | 0.5574 | - | - | | 0.392 | 4900 | 0.4953 | - | - | | 0.4 | 5000 | 0.5365 | 0.5969 | 0.779 | | 0.408 | 5100 | 0.4835 | - | - | | 0.416 | 5200 | 0.4573 | - | - | | 0.424 | 5300 | 0.5554 | - | - | | 0.432 | 5400 | 0.5623 | - | - | | 0.44 | 5500 | 0.5955 | - | - | | 0.448 | 5600 | 0.5086 | - | - | | 0.456 | 5700 | 0.5081 | - | - | | 0.464 | 5800 | 0.4829 | - | - | | 0.472 | 5900 | 0.5066 | - | - | | 0.48 | 6000 | 0.4997 | 0.5920 | 0.776 | | 0.488 | 6100 | 0.5075 | - | - | | 0.496 | 6200 | 0.5051 | - | - | | 0.504 | 6300 | 0.5019 | - | - | | 0.512 | 6400 | 0.4774 | - | - | | 0.52 | 6500 | 0.4975 | - | - | | 0.528 | 6600 | 0.4756 | - | - | | 0.536 | 6700 | 0.4656 | - | - | | 0.544 | 6800 | 0.4671 | - | - | | 0.552 | 6900 | 0.4646 | - | - | | 0.56 | 7000 | 0.5595 | 0.5853 | 0.777 | | 0.568 | 7100 | 0.4812 | - | - | | 0.576 | 7200 | 0.506 | - | - | | 0.584 | 7300 | 0.49 | - | - | | 0.592 | 7400 | 0.464 | - | - | | 0.6 | 7500 | 0.441 | - | - | | 0.608 | 7600 | 0.4492 | - | - | | 0.616 | 7700 | 0.457 | - | - | | 0.624 | 7800 | 0.493 | - | - | | 0.632 | 7900 | 0.4174 | - | - | | 0.64 | 8000 | 0.4686 | 0.5809 | 0.785 | | 0.648 | 8100 | 0.4529 | - | - | | 0.656 | 8200 | 0.4784 | - | - | | 0.664 | 8300 | 0.4697 | - | - | | 0.672 | 8400 | 0.4489 | - | - | | 0.68 | 8500 | 0.4439 | - | - | | 0.688 | 8600 | 0.4063 | - | - | | 0.696 | 8700 | 0.4634 | - | - | | 0.704 | 8800 | 0.4446 | - | - | | 0.712 | 8900 | 0.4725 | - | - | | 0.72 | 9000 | 0.3954 | 0.5769 | 0.781 | | 0.728 | 9100 | 0.4536 | - | - | | 0.736 | 9200 | 0.4583 | - | - | | 0.744 | 9300 | 0.4415 | - | - | | 0.752 | 9400 | 0.4716 | - | - | | 0.76 | 9500 | 0.4393 | - | - | | 0.768 | 9600 | 0.4332 | - | - | | 0.776 | 9700 | 0.4236 | - | - | | 0.784 | 9800 | 0.4021 | - | - | | 0.792 | 9900 | 0.4324 | - | - | | 0.8 | 10000 | 0.4197 | 0.5796 | 0.78 | | 0.808 | 10100 | 0.4576 | - | - | | 0.816 | 10200 | 0.4238 | - | - | | 0.824 | 10300 | 0.4468 | - | - | | 0.832 | 10400 | 0.4301 | - | - | | 0.84 | 10500 | 0.414 | - | - | | 0.848 | 10600 | 0.4563 | - | - | | 0.856 | 10700 | 0.4212 | - | - | | 0.864 | 10800 | 0.3905 | - | - | | 0.872 | 10900 | 0.4384 | - | - | | 0.88 | 11000 | 0.3474 | 0.5709 | 0.788 | | 0.888 | 11100 | 0.4396 | - | - | | 0.896 | 11200 | 0.3819 | - | - | | 0.904 | 11300 | 0.3748 | - | - | | 0.912 | 11400 | 0.4217 | - | - | | 0.92 | 11500 | 0.3893 | - | - | | 0.928 | 11600 | 0.3835 | - | - | | 0.936 | 11700 | 0.4303 | - | - | | 0.944 | 11800 | 0.4274 | - | - | | 0.952 | 11900 | 0.4089 | - | - | | 0.96 | 12000 | 0.4009 | 0.5710 | 0.786 | | 0.968 | 12100 | 0.3832 | - | - | | 0.976 | 12200 | 0.3543 | - | - | | 0.984 | 12300 | 0.4866 | - | - | | 0.992 | 12400 | 0.4531 | - | - | | 1.0 | 12500 | 0.3728 | - | - | | 1.008 | 12600 | 0.386 | - | - | | 1.016 | 12700 | 0.3622 | - | - | | 1.024 | 12800 | 0.4013 | - | - | | 1.032 | 12900 | 0.3543 | - | - | | 1.04 | 13000 | 0.3918 | 0.5712 | 0.792 | | 1.048 | 13100 | 0.3961 | - | - | | 1.056 | 13200 | 0.3804 | - | - | | 1.064 | 13300 | 0.4049 | - | - | | 1.072 | 13400 | 0.3374 | - | - | | 1.08 | 13500 | 0.3746 | - | - | | 1.088 | 13600 | 0.3162 | - | - | | 1.096 | 13700 | 0.3536 | - | - | | 1.104 | 13800 | 0.3101 | - | - | | 1.112 | 13900 | 0.3704 | - | - | | 1.12 | 14000 | 0.3412 | 0.5758 | 0.788 | | 1.1280 | 14100 | 0.342 | - | - | | 1.1360 | 14200 | 0.383 | - | - | | 1.144 | 14300 | 0.3554 | - | - | | 1.152 | 14400 | 0.4013 | - | - | | 1.16 | 14500 | 0.3486 | - | - | | 1.168 | 14600 | 0.3367 | - | - | | 1.176 | 14700 | 0.3737 | - | - | | 1.184 | 14800 | 0.319 | - | - | | 1.192 | 14900 | 0.3211 | - | - | | 1.2 | 15000 | 0.3284 | 0.5804 | 0.787 |
### Framework Versions - Python: 3.10.12 - Sentence Transformers: 3.0.0 - Transformers: 4.38.2 - PyTorch: 2.1.2+cu121 - Accelerate: 0.27.2 - Datasets: 2.19.1 - Tokenizers: 0.15.2 ## Citation ### BibTeX #### Sentence Transformers ```bibtex @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", } ``` #### CachedMultipleNegativesRankingLoss ```bibtex @misc{gao2021scaling, title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup}, author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan}, year={2021}, eprint={2101.06983}, archivePrefix={arXiv}, primaryClass={cs.LG} } ```