adriansanz/SITGES-aina4-FT1

BGE SITGES CAT

This is a sentence-transformers model finetuned from projecte-aina/ST-NLI-ca_paraphrase-multilingual-mpnet-base. 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: projecte-aina/ST-NLI-ca_paraphrase-multilingual-mpnet-base
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity
• Language: ca
• 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': 128, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (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/SITGES-aina4_moreseq")
# Run inference
sentences = [
    "Mitjançant aquest tràmit la persona interessada posa en coneixement de l'Ajuntament de Sitges l'inici d'un espectacle públic o activitat recreativa de caràcter extraordinari...",
    'Quin és el paper de la persona interessada en la llicència per a espectacles públics o activitats recreatives de caràcter extraordinari?',
    "Quin és el paper del Registre de Sol·licitants d'Habitatge amb Protecció Oficial en la gestió d'habitatges?",
]
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	0.0733
cosine_accuracy@3	0.1573
cosine_accuracy@5	0.2177
cosine_accuracy@10	0.3944
cosine_precision@1	0.0733
cosine_precision@3	0.0524
cosine_precision@5	0.0435
cosine_precision@10	0.0394
cosine_recall@1	0.0733
cosine_recall@3	0.1573
cosine_recall@5	0.2177
cosine_recall@10	0.3944
cosine_ndcg@10	0.2013
cosine_mrr@10	0.1439
cosine_map@100	0.171

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0733
cosine_accuracy@3	0.1509
cosine_accuracy@5	0.2177
cosine_accuracy@10	0.3944
cosine_precision@1	0.0733
cosine_precision@3	0.0503
cosine_precision@5	0.0435
cosine_precision@10	0.0394
cosine_recall@1	0.0733
cosine_recall@3	0.1509
cosine_recall@5	0.2177
cosine_recall@10	0.3944
cosine_ndcg@10	0.2016
cosine_mrr@10	0.1444
cosine_map@100	0.1716

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0733
cosine_accuracy@3	0.1487
cosine_accuracy@5	0.2112
cosine_accuracy@10	0.4009
cosine_precision@1	0.0733
cosine_precision@3	0.0496
cosine_precision@5	0.0422
cosine_precision@10	0.0401
cosine_recall@1	0.0733
cosine_recall@3	0.1487
cosine_recall@5	0.2112
cosine_recall@10	0.4009
cosine_ndcg@10	0.2021
cosine_mrr@10	0.1434
cosine_map@100	0.1697

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.069
cosine_accuracy@3	0.1466
cosine_accuracy@5	0.2177
cosine_accuracy@10	0.3815
cosine_precision@1	0.069
cosine_precision@3	0.0489
cosine_precision@5	0.0435
cosine_precision@10	0.0381
cosine_recall@1	0.069
cosine_recall@3	0.1466
cosine_recall@5	0.2177
cosine_recall@10	0.3815
cosine_ndcg@10	0.1954
cosine_mrr@10	0.1398
cosine_map@100	0.166

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.056
cosine_accuracy@3	0.1379
cosine_accuracy@5	0.194
cosine_accuracy@10	0.3685
cosine_precision@1	0.056
cosine_precision@3	0.046
cosine_precision@5	0.0388
cosine_precision@10	0.0369
cosine_recall@1	0.056
cosine_recall@3	0.1379
cosine_recall@5	0.194
cosine_recall@10	0.3685
cosine_ndcg@10	0.1823
cosine_mrr@10	0.1269
cosine_map@100	0.1543

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: 6
• 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: 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: 6
• 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	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
0.3065	5	3.3947	-	-	-	-	-	-
0.6130	10	2.6401	-	-	-	-	-	-
0.9195	15	2.0152	-	-	-	-	-	-
0.9808	16	-	1.3404	0.1639	0.1577	0.1694	0.1503	0.1638
1.2261	20	1.4542	-	-	-	-	-	-
1.5326	25	1.0135	-	-	-	-	-	-
1.8391	30	0.8437	-	-	-	-	-	-
1.9617	32	-	0.9436	0.1556	0.1596	0.1600	0.1467	0.1701
2.1456	35	0.7676	-	-	-	-	-	-
2.4521	40	0.5126	-	-	-	-	-	-
2.7586	45	0.4358	-	-	-	-	-	-
2.9425	48	-	0.7852	0.1650	0.1693	0.1720	0.1511	0.1686
3.0651	50	0.4192	-	-	-	-	-	-
3.3716	55	0.3429	-	-	-	-	-	-
3.6782	60	0.3025	-	-	-	-	-	-
3.9847	65	0.2863	0.7401	0.1646	0.1706	0.1759	0.1480	0.1694
4.2912	70	0.2474	-	-	-	-	-	-
4.5977	75	0.2324	-	-	-	-	-	-
4.9042	80	0.2344	-	-	-	-	-	-
4.9655	81	-	0.7217	0.1663	0.1699	0.1767	0.1512	0.1696
5.2107	85	0.2181	-	-	-	-	-	-
5.5172	90	0.2116	-	-	-	-	-	-
5.8238	95	0.1926	-	-	-	-	-	-
5.8851	96	-	0.7154	0.166	0.1697	0.1716	0.1543	0.171

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.42.3
• PyTorch: 2.3.1+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}
}