SentenceTransformer based on BAAI/bge-m3
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
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/sitgrsBAAIbge-m3-300824")
# Run inference
sentences = [
'Per valorar l’interès de la proposta es tindrà en compte: Tipus d’activitat Antecedents Dates de celebració Accions de promoció dutes a terme des de l’organització Nivell de molèstia previst i interferència en la vida quotidiana.',
"Quin és el paper de les accions de promoció en les subvencions per a projectes i activitats de l'àmbit turístic?",
"Quin és el benefici de la realització d'exposicions al Centre Cultural Miramar?",
]
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 | 0.0591 |
cosine_accuracy@3 | 0.1276 |
cosine_accuracy@5 | 0.1735 |
cosine_accuracy@10 | 0.2861 |
cosine_precision@1 | 0.0591 |
cosine_precision@3 | 0.0425 |
cosine_precision@5 | 0.0347 |
cosine_precision@10 | 0.0286 |
cosine_recall@1 | 0.0591 |
cosine_recall@3 | 0.1276 |
cosine_recall@5 | 0.1735 |
cosine_recall@10 | 0.2861 |
cosine_ndcg@10 | 0.1537 |
cosine_mrr@10 | 0.1139 |
cosine_map@100 | 0.1398 |
Information Retrieval
- Dataset:
dim_768
- Evaluated with
InformationRetrievalEvaluator
Metric | Value |
---|---|
cosine_accuracy@1 | 0.0591 |
cosine_accuracy@3 | 0.1257 |
cosine_accuracy@5 | 0.1801 |
cosine_accuracy@10 | 0.2946 |
cosine_precision@1 | 0.0591 |
cosine_precision@3 | 0.0419 |
cosine_precision@5 | 0.036 |
cosine_precision@10 | 0.0295 |
cosine_recall@1 | 0.0591 |
cosine_recall@3 | 0.1257 |
cosine_recall@5 | 0.1801 |
cosine_recall@10 | 0.2946 |
cosine_ndcg@10 | 0.1564 |
cosine_mrr@10 | 0.1149 |
cosine_map@100 | 0.1405 |
Information Retrieval
- Dataset:
dim_512
- Evaluated with
InformationRetrievalEvaluator
Metric | Value |
---|---|
cosine_accuracy@1 | 0.0591 |
cosine_accuracy@3 | 0.1257 |
cosine_accuracy@5 | 0.1707 |
cosine_accuracy@10 | 0.2983 |
cosine_precision@1 | 0.0591 |
cosine_precision@3 | 0.0419 |
cosine_precision@5 | 0.0341 |
cosine_precision@10 | 0.0298 |
cosine_recall@1 | 0.0591 |
cosine_recall@3 | 0.1257 |
cosine_recall@5 | 0.1707 |
cosine_recall@10 | 0.2983 |
cosine_ndcg@10 | 0.1571 |
cosine_mrr@10 | 0.115 |
cosine_map@100 | 0.1397 |
Information Retrieval
- Dataset:
dim_256
- Evaluated with
InformationRetrievalEvaluator
Metric | Value |
---|---|
cosine_accuracy@1 | 0.0516 |
cosine_accuracy@3 | 0.121 |
cosine_accuracy@5 | 0.1679 |
cosine_accuracy@10 | 0.2889 |
cosine_precision@1 | 0.0516 |
cosine_precision@3 | 0.0403 |
cosine_precision@5 | 0.0336 |
cosine_precision@10 | 0.0289 |
cosine_recall@1 | 0.0516 |
cosine_recall@3 | 0.121 |
cosine_recall@5 | 0.1679 |
cosine_recall@10 | 0.2889 |
cosine_ndcg@10 | 0.1498 |
cosine_mrr@10 | 0.1082 |
cosine_map@100 | 0.1338 |
Information Retrieval
- Dataset:
dim_128
- Evaluated with
InformationRetrievalEvaluator
Metric | Value |
---|---|
cosine_accuracy@1 | 0.0516 |
cosine_accuracy@3 | 0.1173 |
cosine_accuracy@5 | 0.1717 |
cosine_accuracy@10 | 0.2889 |
cosine_precision@1 | 0.0516 |
cosine_precision@3 | 0.0391 |
cosine_precision@5 | 0.0343 |
cosine_precision@10 | 0.0289 |
cosine_recall@1 | 0.0516 |
cosine_recall@3 | 0.1173 |
cosine_recall@5 | 0.1717 |
cosine_recall@10 | 0.2889 |
cosine_ndcg@10 | 0.1488 |
cosine_mrr@10 | 0.1069 |
cosine_map@100 | 0.1328 |
Information Retrieval
- Dataset:
dim_64
- Evaluated with
InformationRetrievalEvaluator
Metric | Value |
---|---|
cosine_accuracy@1 | 0.0507 |
cosine_accuracy@3 | 0.1126 |
cosine_accuracy@5 | 0.1642 |
cosine_accuracy@10 | 0.2824 |
cosine_precision@1 | 0.0507 |
cosine_precision@3 | 0.0375 |
cosine_precision@5 | 0.0328 |
cosine_precision@10 | 0.0282 |
cosine_recall@1 | 0.0507 |
cosine_recall@3 | 0.1126 |
cosine_recall@5 | 0.1642 |
cosine_recall@10 | 0.2824 |
cosine_ndcg@10 | 0.1449 |
cosine_mrr@10 | 0.104 |
cosine_map@100 | 0.1306 |
Training Details
Training Dataset
Unnamed Dataset
- Size: 9,593 training samples
- Columns:
positive
andanchor
- Approximate statistics based on the first 1000 samples:
positive anchor type string string details - min: 5 tokens
- mean: 49.28 tokens
- max: 178 tokens
- min: 10 tokens
- mean: 21.16 tokens
- max: 41 tokens
- Samples:
positive anchor Mitjançant aquest tràmit la persona interessada posa en coneixement de l'Ajuntament l’inici o modificació substancial d’una activitat econòmica, i hi adjunta el certificat tècnic acreditatiu del compliment dels requisits necessaris que estableix la normativa vigent per a l‘exercici de l’activitat.
Quin és el resultat esperat després de presentar el certificat tècnic en el tràmit de comunicació d'inici d'activitat?
L'Ajuntament de Sitges ofereix a aquelles famílies que acompleixin els requisits establerts, ajuts per al pagament de la quota del servei i de la quota del menjador dels infants matriculats a les Llars d'Infants Municipals ( 0-3 anys).
Quins són els requisits per a beneficiar-se dels ajuts de l'Ajuntament de Sitges?
Les entitats o associacions culturals han de presentar la sol·licitud de subvenció dins del termini establert per l'Ajuntament de Sitges.
Quin és el termini per a presentar una sol·licitud de subvenció per a un projecte cultural?
- 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
: epochper_device_train_batch_size
: 16per_device_eval_batch_size
: 16gradient_accumulation_steps
: 16learning_rate
: 2e-05num_train_epochs
: 5lr_scheduler_type
: cosinewarmup_ratio
: 0.2bf16
: Truetf32
: Trueload_best_model_at_end
: Trueoptim
: adamw_torch_fusedbatch_sampler
: no_duplicates
All Hyperparameters
Click to expand
overwrite_output_dir
: Falsedo_predict
: Falseeval_strategy
: epochprediction_loss_only
: Trueper_device_train_batch_size
: 16per_device_eval_batch_size
: 16per_gpu_train_batch_size
: Noneper_gpu_eval_batch_size
: Nonegradient_accumulation_steps
: 16eval_accumulation_steps
: Nonetorch_empty_cache_steps
: Nonelearning_rate
: 2e-05weight_decay
: 0.0adam_beta1
: 0.9adam_beta2
: 0.999adam_epsilon
: 1e-08max_grad_norm
: 1.0num_train_epochs
: 5max_steps
: -1lr_scheduler_type
: cosinelr_scheduler_kwargs
: {}warmup_ratio
: 0.2warmup_steps
: 0log_level
: passivelog_level_replica
: warninglog_on_each_node
: Truelogging_nan_inf_filter
: Truesave_safetensors
: Truesave_on_each_node
: Falsesave_only_model
: Falserestore_callback_states_from_checkpoint
: Falseno_cuda
: Falseuse_cpu
: Falseuse_mps_device
: Falseseed
: 42data_seed
: Nonejit_mode_eval
: Falseuse_ipex
: Falsebf16
: Truefp16
: Falsefp16_opt_level
: O1half_precision_backend
: autobf16_full_eval
: Falsefp16_full_eval
: Falsetf32
: Truelocal_rank
: 0ddp_backend
: Nonetpu_num_cores
: Nonetpu_metrics_debug
: Falsedebug
: []dataloader_drop_last
: Falsedataloader_num_workers
: 0dataloader_prefetch_factor
: Nonepast_index
: -1disable_tqdm
: Falseremove_unused_columns
: Truelabel_names
: Noneload_best_model_at_end
: Trueignore_data_skip
: Falsefsdp
: []fsdp_min_num_params
: 0fsdp_config
: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}fsdp_transformer_layer_cls_to_wrap
: Noneaccelerator_config
: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}deepspeed
: Nonelabel_smoothing_factor
: 0.0optim
: adamw_torch_fusedoptim_args
: Noneadafactor
: Falsegroup_by_length
: Falselength_column_name
: lengthddp_find_unused_parameters
: Noneddp_bucket_cap_mb
: Noneddp_broadcast_buffers
: Falsedataloader_pin_memory
: Truedataloader_persistent_workers
: Falseskip_memory_metrics
: Trueuse_legacy_prediction_loop
: Falsepush_to_hub
: Falseresume_from_checkpoint
: Nonehub_model_id
: Nonehub_strategy
: every_savehub_private_repo
: Falsehub_always_push
: Falsegradient_checkpointing
: Falsegradient_checkpointing_kwargs
: Noneinclude_inputs_for_metrics
: Falseeval_do_concat_batches
: Truefp16_backend
: autopush_to_hub_model_id
: Nonepush_to_hub_organization
: Nonemp_parameters
:auto_find_batch_size
: Falsefull_determinism
: Falsetorchdynamo
: Noneray_scope
: lastddp_timeout
: 1800torch_compile
: Falsetorch_compile_backend
: Nonetorch_compile_mode
: Nonedispatch_batches
: Nonesplit_batches
: Noneinclude_tokens_per_second
: Falseinclude_num_input_tokens_seen
: Falseneftune_noise_alpha
: Noneoptim_target_modules
: Nonebatch_eval_metrics
: Falseeval_on_start
: Falseeval_use_gather_object
: Falsebatch_sampler
: no_duplicatesmulti_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 |
---|---|---|---|---|---|---|---|---|
0.2667 | 10 | 3.5318 | - | - | - | - | - | - |
0.5333 | 20 | 2.3744 | - | - | - | - | - | - |
0.8 | 30 | 1.6587 | - | - | - | - | - | - |
0.9867 | 37 | - | 0.1350 | 0.1317 | 0.1349 | 0.1341 | 0.1207 | 0.1322 |
1.0667 | 40 | 1.1513 | - | - | - | - | - | - |
1.3333 | 50 | 1.0055 | - | - | - | - | - | - |
1.6 | 60 | 0.7369 | - | - | - | - | - | - |
1.8667 | 70 | 0.4855 | - | - | - | - | - | - |
2.0 | 75 | - | 0.1366 | 0.1370 | 0.1376 | 0.1345 | 0.1290 | 0.1355 |
2.1333 | 80 | 0.4362 | - | - | - | - | - | - |
2.4 | 90 | 0.3943 | - | - | - | - | - | - |
2.6667 | 100 | 0.3495 | - | - | - | - | - | - |
2.9333 | 110 | 0.2138 | - | - | - | - | - | - |
2.9867 | 112 | - | 0.1364 | 0.1342 | 0.1374 | 0.1361 | 0.1256 | 0.1367 |
3.2 | 120 | 0.2176 | - | - | - | - | - | - |
3.4667 | 130 | 0.2513 | - | - | - | - | - | - |
3.7333 | 140 | 0.2163 | - | - | - | - | - | - |
4.0 | 150 | 0.15 | 0.1401 | 0.1308 | 0.1332 | 0.1396 | 0.1279 | 0.1396 |
4.2667 | 160 | 0.1613 | - | - | - | - | - | - |
4.5333 | 170 | 0.1955 | - | - | - | - | - | - |
4.8 | 180 | 0.1514 | - | - | - | - | - | - |
4.9333 | 185 | - | 0.1398 | 0.1328 | 0.1338 | 0.1397 | 0.1306 | 0.1405 |
- The bold row denotes the saved checkpoint.
Framework Versions
- Python: 3.10.12
- Sentence Transformers: 3.0.1
- Transformers: 4.44.2
- 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}
}
- Downloads last month
- 100
Inference API (serverless) is not available, repository is disabled.
Model tree for adriansanz/sitges-v1
Base model
BAAI/bge-m3
Finetuned
this model
Evaluation results
- Cosine Accuracy@1 on dim 1024self-reported0.059
- Cosine Accuracy@3 on dim 1024self-reported0.128
- Cosine Accuracy@5 on dim 1024self-reported0.174
- Cosine Accuracy@10 on dim 1024self-reported0.286
- Cosine Precision@1 on dim 1024self-reported0.059
- Cosine Precision@3 on dim 1024self-reported0.043
- Cosine Precision@5 on dim 1024self-reported0.035
- Cosine Precision@10 on dim 1024self-reported0.029
- Cosine Recall@1 on dim 1024self-reported0.059
- Cosine Recall@3 on dim 1024self-reported0.128