adriansanz/ST-tramits-sitges-003-10ep

SentenceTransformer based on BAAI/bge-m3

This is a sentence-transformers model finetuned from BAAI/bge-m3 on the json dataset. 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
• Training Dataset:
  • json

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("adriansanz/ST-tramits-sitges-003-10ep")
# Run inference
sentences = [
    "Els comerços locals obtenen un benefici principal de la implementació del projecte d'implantació i ús de la targeta de fidelització del comerç local de Sitges, que és la possibilitat d'augmentar les vendes i la fidelització dels clients.",
    "Quin és el benefici que els comerços locals obtenen de la implementació del projecte d'implantació i ús de la targeta de fidelització?",
    'Quin és el propòsit de la deixalleria municipal per a l’ambient?',
]
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	0.1331
cosine_accuracy@3	0.2624
cosine_accuracy@5	0.3536
cosine_accuracy@10	0.5243
cosine_precision@1	0.1331
cosine_precision@3	0.0875
cosine_precision@5	0.0707
cosine_precision@10	0.0524
cosine_recall@1	0.1331
cosine_recall@3	0.2624
cosine_recall@5	0.3536
cosine_recall@10	0.5243
cosine_ndcg@10	0.2986
cosine_mrr@10	0.2301
cosine_map@100	0.2513

Information Retrieval

• Dataset: dim_768
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1322
cosine_accuracy@3	0.263
cosine_accuracy@5	0.3541
cosine_accuracy@10	0.5286
cosine_precision@1	0.1322
cosine_precision@3	0.0877
cosine_precision@5	0.0708
cosine_precision@10	0.0529
cosine_recall@1	0.1322
cosine_recall@3	0.263
cosine_recall@5	0.3541
cosine_recall@10	0.5286
cosine_ndcg@10	0.3011
cosine_mrr@10	0.2322
cosine_map@100	0.253

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1342
cosine_accuracy@3	0.2655
cosine_accuracy@5	0.3589
cosine_accuracy@10	0.5257
cosine_precision@1	0.1342
cosine_precision@3	0.0885
cosine_precision@5	0.0718
cosine_precision@10	0.0526
cosine_recall@1	0.1342
cosine_recall@3	0.2655
cosine_recall@5	0.3589
cosine_recall@10	0.5257
cosine_ndcg@10	0.3011
cosine_mrr@10	0.2329
cosine_map@100	0.2538

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1266
cosine_accuracy@3	0.2633
cosine_accuracy@5	0.3564
cosine_accuracy@10	0.5229
cosine_precision@1	0.1266
cosine_precision@3	0.0878
cosine_precision@5	0.0713
cosine_precision@10	0.0523
cosine_recall@1	0.1266
cosine_recall@3	0.2633
cosine_recall@5	0.3564
cosine_recall@10	0.5229
cosine_ndcg@10	0.2972
cosine_mrr@10	0.2285
cosine_map@100	0.2496

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1274
cosine_accuracy@3	0.2684
cosine_accuracy@5	0.3553
cosine_accuracy@10	0.521
cosine_precision@1	0.1274
cosine_precision@3	0.0895
cosine_precision@5	0.0711
cosine_precision@10	0.0521
cosine_recall@1	0.1274
cosine_recall@3	0.2684
cosine_recall@5	0.3553
cosine_recall@10	0.521
cosine_ndcg@10	0.2973
cosine_mrr@10	0.2293
cosine_map@100	0.2507

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1224
cosine_accuracy@3	0.2546
cosine_accuracy@5	0.344
cosine_accuracy@10	0.5165
cosine_precision@1	0.1224
cosine_precision@3	0.0849
cosine_precision@5	0.0688
cosine_precision@10	0.0516
cosine_recall@1	0.1224
cosine_recall@3	0.2546
cosine_recall@5	0.344
cosine_recall@10	0.5165
cosine_ndcg@10	0.2909
cosine_mrr@10	0.2225
cosine_map@100	0.2429

Training Details

Training Dataset

json

• Dataset: json
• Size: 6,399 training samples
• Columns: positive and anchor
• Approximate statistics based on the first 1000 samples:

  	positive	anchor
  type	string	string
  details	min: 9 tokens mean: 49.44 tokens max: 178 tokens	min: 9 tokens mean: 21.17 tokens max: 48 tokens

• Samples:

  positive	anchor
  L'Ajuntament de Sitges atorga subvencions per a projectes i activitats d'interès públic o social que tinguin per finalitat les activitats esportives federades, escolars o populars desenvolupades per les entitats esportives i esportistes del municipi de Sitges.	Quin és el benefici de les subvencions per a les entitats esportives?
  L'Ajuntament de Sitges atorga subvencions per a projectes i activitats d'interès públic o social que tinguin per finalitat les activitats esportives federades, escolars o populars desenvolupades per les entitats esportives i esportistes del municipi de Sitges al llarg de l'exercici per la qual es sol·licita la subvenció, i reuneixin les condicions assenyalades a les bases.	Quin és el període d'execució dels projectes i activitats esportives?
  Certificat on s'indica el nombre d'habitatges que configuren el padró de l'Impost sobre Béns Immobles del municipi o bé d'una part d'aquest.	Quin és el contingut del certificat del nombre d'habitatges?

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

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: 10
• lr_scheduler_type: cosine
• warmup_ratio: 0.2
• 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
• torch_empty_cache_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: 10
• max_steps: -1
• lr_scheduler_type: cosine
• 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: 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
• eval_use_gather_object: False
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training 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
0.4	10	3.5464	-	-	-	-	-	-
0.8	20	2.3861	-	-	-	-	-	-
1.0	25	-	0.2327	0.2144	0.2252	0.2286	0.1938	0.2329
1.1975	30	1.8712	-	-	-	-	-	-
1.5975	40	1.3322	-	-	-	-	-	-
1.9975	50	0.9412	0.2410	0.2310	0.2383	0.2415	0.2236	0.2436
2.395	60	0.806	-	-	-	-	-	-
2.795	70	0.5024	-	-	-	-	-	-
2.995	75	-	0.2451	0.2384	0.2455	0.2487	0.2323	0.2423
3.1925	80	0.4259	-	-	-	-	-	-
3.5925	90	0.3556	-	-	-	-	-	-
3.9925	100	0.2555	0.2477	0.2443	0.2417	0.2485	0.2369	0.2470
4.39	110	0.2611	-	-	-	-	-	-
4.79	120	0.1939	-	-	-	-	-	-
4.99	125	-	0.2490	0.2425	0.2479	0.2485	0.2386	0.2495
5.1875	130	0.2021	-	-	-	-	-	-
5.5875	140	0.1537	-	-	-	-	-	-
5.9875	150	0.1277	0.2535	0.2491	0.2491	0.2534	0.2403	0.2541
6.385	160	0.1213	-	-	-	-	-	-
6.785	170	0.1035	-	-	-	-	-	-
6.985	175	-	0.2513	0.2493	0.2435	0.2515	0.2380	0.2528
7.1825	180	0.0965	-	-	-	-	-	-
7.5825	190	0.0861	-	-	-	-	-	-
7.9825	200	0.0794	0.2529	0.2536	0.2526	0.2545	0.2438	0.2570
8.38	210	0.0734	-	-	-	-	-	-
8.78	220	0.066	-	-	-	-	-	-
8.98	225	-	0.2538	0.2523	0.2519	0.2542	0.2457	0.2572
9.1775	230	0.0731	-	-	-	-	-	-
9.5775	240	0.0726	-	-	-	-	-	-
9.9775	250	0.0632	0.2513	0.2507	0.2496	0.2538	0.2429	0.2530

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.1.1
• Transformers: 4.44.2
• PyTorch: 2.4.1+cu121
• Accelerate: 0.35.0.dev0
• Datasets: 3.0.1
• 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}
}