Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
Paper • 1908.10084 • Published • 12
This is a sentence-transformers model finetuned from BAAI/bge-m3. It maps sentences & paragraphs to a 1024-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
SentenceTransformer(
(0): Transformer({'max_seq_length': 512, 'do_lower_case': False, 'architecture': 'XLMRobertaModel'})
(1): Pooling({'word_embedding_dimension': 1024, '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})
(2): Normalize()
)
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("sentence_transformers_model_id")
# Run inference
sentences = [
"Convention de liquidation du régime matrimonial pendant l'instance en divorce : acte notarié obligatoire pour les biens soumis à publicité foncière ?",
"Code civil - Article 265-2\nDes personnes - Du divorce - Des conséquences du divorce - Des conséquences du divorce pour les époux - Dispositions générales.\nLes époux peuvent, pendant l'instance en divorce, passer toutes conventions pour la liquidation et le partage de leur régime matrimonial.\n Lorsque la liquidation porte sur des biens soumis à la publicité foncière, la convention doit être passée par acte notarié.",
"Code civil - Article 265-1\nLe divorce est sans incidence sur les droits que l'un ou l'autre des époux tient de la loi ou des conventions passées avec des tiers.",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]
# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.7003, 0.4054],
# [0.7003, 1.0000, 0.4614],
# [0.4054, 0.4614, 1.0000]])
notarial-evalInformationRetrievalEvaluator| Metric | Value |
|---|---|
| cosine_accuracy@1 | 0.9717 |
| cosine_accuracy@3 | 0.9957 |
| cosine_accuracy@5 | 0.9983 |
| cosine_accuracy@10 | 0.9983 |
| cosine_precision@1 | 0.9717 |
| cosine_precision@3 | 0.3319 |
| cosine_precision@5 | 0.1997 |
| cosine_precision@10 | 0.0998 |
| cosine_recall@1 | 0.9717 |
| cosine_recall@3 | 0.9957 |
| cosine_recall@5 | 0.9983 |
| cosine_recall@10 | 0.9983 |
| cosine_ndcg@10 | 0.9876 |
| cosine_mrr@10 | 0.9839 |
| cosine_map@100 | 0.9839 |
sentence_0, sentence_1, and sentence_2| sentence_0 | sentence_1 | sentence_2 | |
|---|---|---|---|
| type | string | string | string |
| details |
|
|
|
| sentence_0 | sentence_1 | sentence_2 |
|---|---|---|
Quels sont les éléments obligatoires et les cas de révision de la lettre de mission du commissaire aux comptes ? |
Code de commerce - Article A821-62 |
Code de commerce - Article A821-62 |
Où consulter la liste provisoire des caisses de mutualité sociale agricole ? |
Code rural et de la pêche maritime - Article R723-28 |
Code rural et de la pêche maritime - Article R723-26 |
Comment majorer un actif d'impôt différé lié à une perte qualifiée au taux minimum ? |
Code général des impôts - Article 223 VU quinquies |
Code général des impôts - Article 223 VU ter |
MultipleNegativesRankingLoss with these parameters:{
"scale": 20.0,
"similarity_fct": "cos_sim",
"gather_across_devices": false,
"directions": [
"query_to_doc"
],
"partition_mode": "joint",
"hardness_mode": null,
"hardness_strength": 0.0
}
per_device_train_batch_size: 4per_device_eval_batch_size: 4fp16: Truemulti_dataset_batch_sampler: round_robinoverwrite_output_dir: Falsedo_predict: Falseeval_strategy: noprediction_loss_only: Trueper_device_train_batch_size: 4per_device_eval_batch_size: 4per_gpu_train_batch_size: Noneper_gpu_eval_batch_size: Nonegradient_accumulation_steps: 1eval_accumulation_steps: Nonetorch_empty_cache_steps: Nonelearning_rate: 5e-05weight_decay: 0.0adam_beta1: 0.9adam_beta2: 0.999adam_epsilon: 1e-08max_grad_norm: 1num_train_epochs: 3max_steps: -1lr_scheduler_type: linearlr_scheduler_kwargs: {}warmup_ratio: 0.0warmup_steps: 0log_level: passivelog_level_replica: warninglog_on_each_node: Truelogging_nan_inf_filter: Truesave_safetensors: Truesave_on_each_node: Falsesave_only_model: Falserestore_callback_states_from_checkpoint: Falseno_cuda: Falseuse_cpu: Falseuse_mps_device: Falseseed: 42data_seed: Nonejit_mode_eval: Falsebf16: Falsefp16: Truefp16_opt_level: O1half_precision_backend: autobf16_full_eval: Falsefp16_full_eval: Falsetf32: Nonelocal_rank: 0ddp_backend: Nonetpu_num_cores: Nonetpu_metrics_debug: Falsedebug: []dataloader_drop_last: Falsedataloader_num_workers: 0dataloader_prefetch_factor: Nonepast_index: -1disable_tqdm: Falseremove_unused_columns: Truelabel_names: Noneload_best_model_at_end: Falseignore_data_skip: Falsefsdp: []fsdp_min_num_params: 0fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}fsdp_transformer_layer_cls_to_wrap: Noneaccelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}parallelism_config: Nonedeepspeed: Nonelabel_smoothing_factor: 0.0optim: adamw_torch_fusedoptim_args: Noneadafactor: Falsegroup_by_length: Falselength_column_name: lengthproject: huggingfacetrackio_space_id: trackioddp_find_unused_parameters: Noneddp_bucket_cap_mb: Noneddp_broadcast_buffers: Falsedataloader_pin_memory: Truedataloader_persistent_workers: Falseskip_memory_metrics: Trueuse_legacy_prediction_loop: Falsepush_to_hub: Falseresume_from_checkpoint: Nonehub_model_id: Nonehub_strategy: every_savehub_private_repo: Nonehub_always_push: Falsehub_revision: Nonegradient_checkpointing: Falsegradient_checkpointing_kwargs: Noneinclude_inputs_for_metrics: Falseinclude_for_metrics: []eval_do_concat_batches: Truefp16_backend: autopush_to_hub_model_id: Nonepush_to_hub_organization: Nonemp_parameters: auto_find_batch_size: Falsefull_determinism: Falsetorchdynamo: Noneray_scope: lastddp_timeout: 1800torch_compile: Falsetorch_compile_backend: Nonetorch_compile_mode: Noneinclude_tokens_per_second: Falseinclude_num_input_tokens_seen: noneftune_noise_alpha: Noneoptim_target_modules: Nonebatch_eval_metrics: Falseeval_on_start: Falseuse_liger_kernel: Falseliger_kernel_config: Noneeval_use_gather_object: Falseaverage_tokens_across_devices: Trueprompts: Nonebatch_sampler: batch_samplermulti_dataset_batch_sampler: round_robinrouter_mapping: {}learning_rate_mapping: {}| Epoch | Step | Training Loss | notarial-eval_cosine_ndcg@10 |
|---|---|---|---|
| 0.0576 | 500 | 0.1604 | - |
| 0.1151 | 1000 | 0.113 | - |
| 0.1727 | 1500 | 0.0933 | - |
| 0.2303 | 2000 | 0.0985 | - |
| 0.2878 | 2500 | 0.1137 | - |
| 0.3454 | 3000 | 0.128 | - |
| 0.4029 | 3500 | 0.1042 | - |
| 0.4605 | 4000 | 0.1132 | - |
| 0.5181 | 4500 | 0.1289 | - |
| 0.5756 | 5000 | 0.1185 | - |
| 0.6332 | 5500 | 0.1119 | - |
| 0.6908 | 6000 | 0.1013 | - |
| 0.7483 | 6500 | 0.1081 | - |
| 0.8059 | 7000 | 0.1237 | - |
| 0.8635 | 7500 | 0.0925 | - |
| 0.9210 | 8000 | 0.0998 | - |
| 0.9786 | 8500 | 0.0953 | - |
| 1.0 | 8686 | - | 0.9876 |
@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",
}
@misc{oord2019representationlearningcontrastivepredictive,
title={Representation Learning with Contrastive Predictive Coding},
author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
year={2019},
eprint={1807.03748},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/1807.03748},
}
Base model
BAAI/bge-m3