Edit model card

BGE base Financial Matryoshka

This is a sentence-transformers model finetuned from BAAI/bge-base-en-v1.5. 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: BAAI/bge-base-en-v1.5
  • Maximum Sequence Length: 512 tokens
  • Output Dimensionality: 768 tokens
  • Similarity Function: Cosine Similarity
  • Language: en
  • License: apache-2.0

Model Sources

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': True}) 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})
  (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("adriansanz/sitges2608")
# Run inference
sentences = [
    'El termini per a la presentació de les sol·licituds de modificació del projecte o activitat subvencionat és de 15 dies naturals abans de la finalització del projecte o activitat.',
    'Quin és el termini per a la presentació de les sol·licituds de modificació del projecte o activitat subvencionat?',
    "Quin és el registre on es troben les dades d'inscripció?",
]
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

Information Retrieval

Metric Value
cosine_accuracy@1 0.0625
cosine_accuracy@3 0.1164
cosine_accuracy@5 0.181
cosine_accuracy@10 0.3556
cosine_precision@1 0.0625
cosine_precision@3 0.0388
cosine_precision@5 0.0362
cosine_precision@10 0.0356
cosine_recall@1 0.0625
cosine_recall@3 0.1164
cosine_recall@5 0.181
cosine_recall@10 0.3556
cosine_ndcg@10 0.1755
cosine_mrr@10 0.1225
cosine_map@100 0.1488

Information Retrieval

Metric Value
cosine_accuracy@1 0.0625
cosine_accuracy@3 0.1099
cosine_accuracy@5 0.1703
cosine_accuracy@10 0.3556
cosine_precision@1 0.0625
cosine_precision@3 0.0366
cosine_precision@5 0.0341
cosine_precision@10 0.0356
cosine_recall@1 0.0625
cosine_recall@3 0.1099
cosine_recall@5 0.1703
cosine_recall@10 0.3556
cosine_ndcg@10 0.1728
cosine_mrr@10 0.1193
cosine_map@100 0.1455

Information Retrieval

Metric Value
cosine_accuracy@1 0.056
cosine_accuracy@3 0.1228
cosine_accuracy@5 0.1724
cosine_accuracy@10 0.3405
cosine_precision@1 0.056
cosine_precision@3 0.0409
cosine_precision@5 0.0345
cosine_precision@10 0.0341
cosine_recall@1 0.056
cosine_recall@3 0.1228
cosine_recall@5 0.1724
cosine_recall@10 0.3405
cosine_ndcg@10 0.168
cosine_mrr@10 0.1168
cosine_map@100 0.1431

Information Retrieval

Metric Value
cosine_accuracy@1 0.0517
cosine_accuracy@3 0.1142
cosine_accuracy@5 0.1832
cosine_accuracy@10 0.319
cosine_precision@1 0.0517
cosine_precision@3 0.0381
cosine_precision@5 0.0366
cosine_precision@10 0.0319
cosine_recall@1 0.0517
cosine_recall@3 0.1142
cosine_recall@5 0.1832
cosine_recall@10 0.319
cosine_ndcg@10 0.1589
cosine_mrr@10 0.1112
cosine_map@100 0.1376

Information Retrieval

Metric Value
cosine_accuracy@1 0.0453
cosine_accuracy@3 0.1056
cosine_accuracy@5 0.1659
cosine_accuracy@10 0.306
cosine_precision@1 0.0453
cosine_precision@3 0.0352
cosine_precision@5 0.0332
cosine_precision@10 0.0306
cosine_recall@1 0.0453
cosine_recall@3 0.1056
cosine_recall@5 0.1659
cosine_recall@10 0.306
cosine_ndcg@10 0.149
cosine_mrr@10 0.1024
cosine_map@100 0.1259

Training Details

Training Dataset

Unnamed Dataset

  • Size: 4,173 training samples
  • Columns: positive and anchor
  • Approximate statistics based on the first 1000 samples:
    positive anchor
    type string string
    details
    • min: 8 tokens
    • mean: 66.25 tokens
    • max: 165 tokens
    • min: 12 tokens
    • mean: 28.12 tokens
    • max: 62 tokens
  • Samples:
    positive anchor
    La persona titular d'una llicència de vehicle lleuger per al servei públic (auto-taxi), en produïr-se un canvi de vehicle, ha de notificar a l'Ajuntament les dades del nou vehicle. Quin és el propòsit de la notificació de les dades del nou vehicle?
    S'entén per garantia l'ingrés a la Tresoreria de l'Ajuntament d'una quantitat econòmica que garanteix el compliment d'una obligació adquirida amb aquest (garanties de concursos o licitacions, fraccionaments de tributs en via executiva, reposició de paviments per obres, etc.). Què s'entén per garantia a l'Ajuntament de Sitges?
    L'ús d'espais del Centre Cultural Miramar per a la realització d'exposicions. Quin és el centre cultural on es poden realitzar les exposicions d'art?
  • Loss: MatryoshkaLoss with these parameters:
    {
        "loss": "MultipleNegativesRankingLoss",
        "matryoshka_dims": [
            768,
            512,
            256,
            128,
            64
        ],
        "matryoshka_weights": [
            1,
            1,
            1,
            1,
            1
        ],
        "n_dims_per_step": -1
    }
    

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: epoch
  • per_device_train_batch_size: 32
  • per_device_eval_batch_size: 16
  • gradient_accumulation_steps: 16
  • learning_rate: 2e-05
  • num_train_epochs: 4
  • lr_scheduler_type: cosine
  • warmup_ratio: 0.1
  • bf16: True
  • tf32: False
  • 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: 32
  • 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: 4
  • 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: False
  • 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 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
0.9771 8 - 0.1210 0.1384 0.1341 0.1002 0.1376
1.2137 10 7.5469 - - - - -
1.9466 16 - 0.136 0.1404 0.1443 0.1249 0.1414
2.4275 20 4.0024 - - - - -
2.9160 24 - 0.1388 0.1460 0.1446 0.1278 0.1436
3.6412 30 3.2149 - - - - -
3.8855 32 - 0.1376 0.1431 0.1455 0.1259 0.1488
  • The bold row denotes the saved checkpoint.

Framework Versions

  • Python: 3.10.12
  • Sentence Transformers: 3.0.1
  • Transformers: 4.42.4
  • PyTorch: 2.3.1+cu121
  • Accelerate: 0.34.0.dev0
  • Datasets: 2.21.0
  • 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}
}
Downloads last month
41
Safetensors
Model size
109M params
Tensor type
F32
·
Inference Examples
Inference API (serverless) is not available, repository is disabled.

Model tree for adriansanz/sitges2608

Finetuned
this model

Evaluation results