adriansanz/sitges10242608-4ep-rerankv4-sp

SentenceTransformer based on PlanTL-GOB-ES/roberta-base-bne

This is a sentence-transformers model finetuned from PlanTL-GOB-ES/roberta-base-bne. 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: PlanTL-GOB-ES/roberta-base-bne
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity

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': 512, 'do_lower_case': False}) with Transformer model: RobertaModel 
  (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:

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/sitges10242608-4ep-rerankv4-sp")
# Run inference
sentences = [
    'Els establiments locals tenen un paper clau en el projecte de la targeta de fidelització, ja que són els que ofereixen descomptes i ofertes especials als consumidors que utilitzen la targeta.',
    'Quin és el paper dels establiments locals en el projecte de la targeta de fidelització?',
    "Quins són els tractaments que beneficien la salut de l'empleat municipal que s'inclouen en l'ajuda?",
]
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

• Dataset: dim_768
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.056
cosine_accuracy@3	0.125
cosine_accuracy@5	0.2134
cosine_accuracy@10	0.4095
cosine_precision@1	0.056
cosine_precision@3	0.0417
cosine_precision@5	0.0427
cosine_precision@10	0.0409
cosine_recall@1	0.056
cosine_recall@3	0.125
cosine_recall@5	0.2134
cosine_recall@10	0.4095
cosine_ndcg@10	0.1939
cosine_mrr@10	0.1301
cosine_map@100	0.1554

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0517
cosine_accuracy@3	0.1228
cosine_accuracy@5	0.2004
cosine_accuracy@10	0.4073
cosine_precision@1	0.0517
cosine_precision@3	0.0409
cosine_precision@5	0.0401
cosine_precision@10	0.0407
cosine_recall@1	0.0517
cosine_recall@3	0.1228
cosine_recall@5	0.2004
cosine_recall@10	0.4073
cosine_ndcg@10	0.1908
cosine_mrr@10	0.1267
cosine_map@100	0.1522

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0582
cosine_accuracy@3	0.1207
cosine_accuracy@5	0.2069
cosine_accuracy@10	0.4159
cosine_precision@1	0.0582
cosine_precision@3	0.0402
cosine_precision@5	0.0414
cosine_precision@10	0.0416
cosine_recall@1	0.0582
cosine_recall@3	0.1207
cosine_recall@5	0.2069
cosine_recall@10	0.4159
cosine_ndcg@10	0.1972
cosine_mrr@10	0.1326
cosine_map@100	0.158

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.056
cosine_accuracy@3	0.1185
cosine_accuracy@5	0.194
cosine_accuracy@10	0.4203
cosine_precision@1	0.056
cosine_precision@3	0.0395
cosine_precision@5	0.0388
cosine_precision@10	0.042
cosine_recall@1	0.056
cosine_recall@3	0.1185
cosine_recall@5	0.194
cosine_recall@10	0.4203
cosine_ndcg@10	0.1948
cosine_mrr@10	0.1286
cosine_map@100	0.1533

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0517
cosine_accuracy@3	0.1336
cosine_accuracy@5	0.2091
cosine_accuracy@10	0.3944
cosine_precision@1	0.0517
cosine_precision@3	0.0445
cosine_precision@5	0.0418
cosine_precision@10	0.0394
cosine_recall@1	0.0517
cosine_recall@3	0.1336
cosine_recall@5	0.2091
cosine_recall@10	0.3944
cosine_ndcg@10	0.1883
cosine_mrr@10	0.1268
cosine_map@100	0.1528

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: 10 tokens mean: 60.84 tokens max: 206 tokens	min: 10 tokens mean: 25.34 tokens max: 53 tokens

• Samples:

  positive	anchor
  L'objectiu principal de la persona coordinadora de colònia felina és garantir el benestar dels animals de la colònia.	Quin és l'objectiu principal de la persona coordinadora de colònia felina?
  Es tracta d'una sala amb capacitat per a 125 persones, equipada amb un petit escenari, sistema de sonorització, pantalla per a projeccions, camerins i serveis higiènics (WC).	Quin és el nombre de persones que pot acollir la sala d'actes del Casal Municipal de la Gent Gran de Sitges?
  Aquest ajut pretén fomentar l’associacionisme empresarial local, per tal de disposar d’agrupacions, gremis o associacions representatives de l’activitat empresarial del municipi.	Quin és el paper de les empreses en aquest ajut?

• 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: 16
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 16
• num_train_epochs: 10
• lr_scheduler_type: cosine
• warmup_ratio: 0.2
• 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: 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: 5e-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: 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
0.6130	10	10.8464	-	-	-	-	-
0.9808	16	-	0.1060	0.1088	0.1067	0.0984	0.1074
1.2261	20	3.5261	-	-	-	-	-
1.8391	30	1.4363	-	-	-	-	-
1.9617	32	-	0.1406	0.1468	0.1356	0.1395	0.1373
2.4521	40	0.5627	-	-	-	-	-
2.9425	48	-	0.1377	0.1418	0.1427	0.1322	0.1437
3.0651	50	0.2727	-	-	-	-	-
3.6782	60	0.1297	-	-	-	-	-
3.9234	64	-	0.1393	0.1457	0.1390	0.1268	0.1462
0.6130	10	0.096	-	-	-	-	-
0.9808	16	-	0.1458	0.1414	0.1443	0.1369	0.1407
1.2261	20	0.1118	-	-	-	-	-
1.8391	30	0.1335	-	-	-	-	-
1.9617	32	-	0.1486	0.1476	0.1419	0.1489	0.1503
2.4521	40	0.0765	-	-	-	-	-
2.9425	48	-	0.1501	0.1459	0.1424	0.1413	0.1437
3.0651	50	0.1449	-	-	-	-	-
3.6782	60	0.0954	-	-	-	-	-
3.9847	65	-	0.1562	0.1559	0.1517	0.1409	0.1553
4.2912	70	0.0786	-	-	-	-	-
4.9042	80	0.0973	-	-	-	-	-
4.9655	81	-	0.1433	0.1397	0.1459	0.1430	0.1457
5.5172	90	0.0334	-	-	-	-	-
5.9464	97	-	0.1499	0.1482	0.1478	0.1466	0.1503
6.1303	100	0.0278	-	-	-	-	-
6.7433	110	0.0223	-	-	-	-	-
6.9885	114	-	0.1561	0.1532	0.1509	0.1519	0.1547
7.3563	120	0.0137	-	-	-	-	-
7.9693	130	0.0129	0.1525	0.1557	0.1505	0.1570	0.1570
8.5824	140	0.0052	-	-	-	-	-
8.9502	146	-	0.1525	0.1586	0.1493	0.1569	0.1553
9.1954	150	0.0044	-	-	-	-	-
9.8084	160	0.0064	0.1533	0.1580	0.1522	0.1528	0.1554

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.42.4
• PyTorch: 2.4.0+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}
}