dariolopez/bge-m3-es-legal-tmp-4

BGE large Legal Spanish

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.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: BAAI/bge-m3
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 1024 tokens
• Similarity Function: Cosine Similarity
• Language: es
• License: apache-2.0

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': 8192, 'do_lower_case': False}) with Transformer model: 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()
)

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("dariolopez/bge-m3-es-legal-tmp-4")
# Run inference
sentences = [
    'Artículo 6. Definiciones. 1. Discriminación directa e indirecta. b) La discriminación indirecta se produce cuando una disposición, criterio o práctica aparentemente neutros ocasiona o puede ocasionar a una o varias personas una desventaja particular con respecto a otras por razón de las causas previstas en el apartado 1 del artículo 2.',
    '¿Qué se considera discriminación indirecta?',
    '¿Qué tipo de información se considera veraz?',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

• Dataset: dim_1024
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5427
cosine_accuracy@3	0.7988
cosine_accuracy@5	0.8384
cosine_accuracy@10	0.8872
cosine_precision@1	0.5427
cosine_precision@3	0.2663
cosine_precision@5	0.1677
cosine_precision@10	0.0887
cosine_recall@1	0.5427
cosine_recall@3	0.7988
cosine_recall@5	0.8384
cosine_recall@10	0.8872
cosine_ndcg@10	0.7233
cosine_mrr@10	0.6696
cosine_map@100	0.6746

Information Retrieval

• Dataset: dim_768
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5396
cosine_accuracy@3	0.8049
cosine_accuracy@5	0.8445
cosine_accuracy@10	0.8902
cosine_precision@1	0.5396
cosine_precision@3	0.2683
cosine_precision@5	0.1689
cosine_precision@10	0.089
cosine_recall@1	0.5396
cosine_recall@3	0.8049
cosine_recall@5	0.8445
cosine_recall@10	0.8902
cosine_ndcg@10	0.7246
cosine_mrr@10	0.6702
cosine_map@100	0.6749

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5488
cosine_accuracy@3	0.8018
cosine_accuracy@5	0.8354
cosine_accuracy@10	0.8933
cosine_precision@1	0.5488
cosine_precision@3	0.2673
cosine_precision@5	0.1671
cosine_precision@10	0.0893
cosine_recall@1	0.5488
cosine_recall@3	0.8018
cosine_recall@5	0.8354
cosine_recall@10	0.8933
cosine_ndcg@10	0.7304
cosine_mrr@10	0.6771
cosine_map@100	0.6811

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5457
cosine_accuracy@3	0.7774
cosine_accuracy@5	0.8293
cosine_accuracy@10	0.872
cosine_precision@1	0.5457
cosine_precision@3	0.2591
cosine_precision@5	0.1659
cosine_precision@10	0.0872
cosine_recall@1	0.5457
cosine_recall@3	0.7774
cosine_recall@5	0.8293
cosine_recall@10	0.872
cosine_ndcg@10	0.7183
cosine_mrr@10	0.6678
cosine_map@100	0.6733

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5335
cosine_accuracy@3	0.7622
cosine_accuracy@5	0.814
cosine_accuracy@10	0.8659
cosine_precision@1	0.5335
cosine_precision@3	0.2541
cosine_precision@5	0.1628
cosine_precision@10	0.0866
cosine_recall@1	0.5335
cosine_recall@3	0.7622
cosine_recall@5	0.814
cosine_recall@10	0.8659
cosine_ndcg@10	0.708
cosine_mrr@10	0.6563
cosine_map@100	0.6617

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5122
cosine_accuracy@3	0.7317
cosine_accuracy@5	0.7896
cosine_accuracy@10	0.8659
cosine_precision@1	0.5122
cosine_precision@3	0.2439
cosine_precision@5	0.1579
cosine_precision@10	0.0866
cosine_recall@1	0.5122
cosine_recall@3	0.7317
cosine_recall@5	0.7896
cosine_recall@10	0.8659
cosine_ndcg@10	0.6908
cosine_mrr@10	0.6347
cosine_map@100	0.6394

Training Details

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 16
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: True
• tf32: True
• load_best_model_at_end: True
• optim: adamw_torch_fused
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• 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: 16
• eval_accumulation_steps: None
• learning_rate: 2e-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: 16
• 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
• 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: True
• fp16: False
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: True
• 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_fused
• 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
• eval_on_start: False
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	dim_1024_cosine_map@100	dim_128_cosine_map@100	dim_256_cosine_map@100	dim_512_cosine_map@100	dim_64_cosine_map@100	dim_768_cosine_map@100
0.4324	5	1.6932	-	-	-	-	-	-	-
0.8649	10	1.1787	-	-	-	-	-	-	-
0.9514	11	-	0.6685	0.6708	0.6300	0.6676	0.6716	0.5560	0.6781
1.2973	15	1.0084	-	-	-	-	-	-	-
1.7297	20	0.5743	-	-	-	-	-	-	-
1.9892	23	-	0.4458	0.6734	0.6533	0.6773	0.6770	0.6174	0.6657
2.1622	25	0.4435	-	-	-	-	-	-	-
2.5946	30	0.2396	-	-	-	-	-	-	-
2.9405	34	-	0.4239	0.6749	0.6591	0.6725	0.6752	0.6188	0.6784
3.0270	35	0.1568	-	-	-	-	-	-	-
3.4595	40	0.1085	-	-	-	-	-	-	-
3.8919	45	0.0582	-	-	-	-	-	-	-
3.9784	46	-	0.3934	0.6820	0.6594	0.6862	0.6856	0.6293	0.6777
4.3243	50	0.0543	-	-	-	-	-	-	-
4.7568	55	0.0349	-	-	-	-	-	-	-
4.9297	57	-	0.3690	0.6747	0.6582	0.6760	0.6852	0.6375	0.6774
5.1892	60	0.03	-	-	-	-	-	-	-
5.6216	65	0.0228	-	-	-	-	-	-	-
5.9676	69	-	0.362	0.6752	0.6643	0.6784	0.6809	0.6312	0.6799
6.0541	70	0.0183	-	-	-	-	-	-	-
6.4865	75	0.0159	-	-	-	-	-	-	-
6.9189	80	0.0113	0.3608	0.6780	0.6582	0.6769	0.6785	0.6366	0.6769
7.3514	85	0.0107	-	-	-	-	-	-	-
7.7838	90	0.0098	-	-	-	-	-	-	-
7.9568	92	-	0.3307	0.6804	0.6511	0.6774	0.6823	0.6355	0.6747
8.2162	95	0.0084	-	-	-	-	-	-	-
8.6486	100	0.0067	-	-	-	-	-	-	-
8.9946	104	-	0.3387	0.6778	0.6518	0.6751	0.6787	0.6313	0.6693
9.0811	105	0.0074	-	-	-	-	-	-	-
9.5135	110	0.0064	-	-	-	-	-	-	-
9.9459	115	0.0052	0.3222	0.6776	0.6571	0.6745	0.6810	0.6397	0.6722
10.3784	120	0.0058	-	-	-	-	-	-	-
10.8108	125	0.0058	-	-	-	-	-	-	-
10.9838	127	-	0.3325	0.6760	0.6595	0.6714	0.6807	0.6399	0.6729
11.2432	130	0.0052	-	-	-	-	-	-	-
11.6757	135	0.0046	-	-	-	-	-	-	-
11.9351	138	-	0.3366	0.6770	0.6598	0.6730	0.6813	0.6360	0.6733
12.1081	140	0.0053	-	-	-	-	-	-	-
12.5405	145	0.0046	-	-	-	-	-	-	-
12.9730	150	0.0045	0.3263	0.6759	0.6599	0.6743	0.6816	0.6394	0.6759
13.4054	155	0.0044	-	-	-	-	-	-	-
13.8378	160	0.0043	-	-	-	-	-	-	-
13.9243	161	-	0.3231	0.6747	0.6593	0.6729	0.6804	0.6407	0.6746
14.2703	165	0.005	-	-	-	-	-	-	-
14.7027	170	0.004	-	-	-	-	-	-	-
14.9622	173	-	0.3238	0.6743	0.6597	0.6720	0.6828	0.6395	0.6759
15.1351	175	0.005	-	-	-	-	-	-	-
15.2216	176	-	0.3244	0.6746	0.6617	0.6733	0.6811	0.6394	0.6749

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.42.3
• PyTorch: 2.2.0+cu121
• Accelerate: 0.32.1
• Datasets: 2.20.0
• Tokenizers: 0.19.1

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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning}, 
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

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}
}