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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("adriansanz/sitgrsBAAIbge-m3-300824v2")
# Run inference
sentences = [
    "Els comerciants locals han de sol·licitar els ajuts per al projecte de la targeta de fidelització dins del termini establert per l'Ajuntament de Sitges.",
    'Quin és el termini perquè els comerciants locals puguin sol·licitar els ajuts per al projecte de la targeta de fidelització?',
    'Quin és el paper de la persona cuidadora en la gestió de les emergències en la colònia felina?',
]
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	0.06
cosine_accuracy@3	0.1304
cosine_accuracy@5	0.1801
cosine_accuracy@10	0.3283
cosine_precision@1	0.06
cosine_precision@3	0.0435
cosine_precision@5	0.036
cosine_precision@10	0.0328
cosine_recall@1	0.06
cosine_recall@3	0.1304
cosine_recall@5	0.1801
cosine_recall@10	0.3283
cosine_ndcg@10	0.168
cosine_mrr@10	0.1205
cosine_map@100	0.1464

Information Retrieval

Dataset: dim_768
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0591
cosine_accuracy@3	0.122
cosine_accuracy@5	0.1811
cosine_accuracy@10	0.3302
cosine_precision@1	0.0591
cosine_precision@3	0.0407
cosine_precision@5	0.0362
cosine_precision@10	0.033
cosine_recall@1	0.0591
cosine_recall@3	0.122
cosine_recall@5	0.1811
cosine_recall@10	0.3302
cosine_ndcg@10	0.1675
cosine_mrr@10	0.1193
cosine_map@100	0.1454

Information Retrieval

Dataset: dim_512
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0582
cosine_accuracy@3	0.1285
cosine_accuracy@5	0.1904
cosine_accuracy@10	0.3265
cosine_precision@1	0.0582
cosine_precision@3	0.0428
cosine_precision@5	0.0381
cosine_precision@10	0.0326
cosine_recall@1	0.0582
cosine_recall@3	0.1285
cosine_recall@5	0.1904
cosine_recall@10	0.3265
cosine_ndcg@10	0.1674
cosine_mrr@10	0.1199
cosine_map@100	0.1464

Information Retrieval

Dataset: dim_256
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0544
cosine_accuracy@3	0.1191
cosine_accuracy@5	0.182
cosine_accuracy@10	0.3171
cosine_precision@1	0.0544
cosine_precision@3	0.0397
cosine_precision@5	0.0364
cosine_precision@10	0.0317
cosine_recall@1	0.0544
cosine_recall@3	0.1191
cosine_recall@5	0.182
cosine_recall@10	0.3171
cosine_ndcg@10	0.161
cosine_mrr@10	0.1145
cosine_map@100	0.1415

Information Retrieval

Dataset: dim_128
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0544
cosine_accuracy@3	0.122
cosine_accuracy@5	0.182
cosine_accuracy@10	0.3114
cosine_precision@1	0.0544
cosine_precision@3	0.0407
cosine_precision@5	0.0364
cosine_precision@10	0.0311
cosine_recall@1	0.0544
cosine_recall@3	0.122
cosine_recall@5	0.182
cosine_recall@10	0.3114
cosine_ndcg@10	0.1596
cosine_mrr@10	0.1144
cosine_map@100	0.1416

Information Retrieval

Dataset: dim_64
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0507
cosine_accuracy@3	0.1144
cosine_accuracy@5	0.1829
cosine_accuracy@10	0.3077
cosine_precision@1	0.0507
cosine_precision@3	0.0381
cosine_precision@5	0.0366
cosine_precision@10	0.0308
cosine_recall@1	0.0507
cosine_recall@3	0.1144
cosine_recall@5	0.1829
cosine_recall@10	0.3077
cosine_ndcg@10	0.1559
cosine_mrr@10	0.1105
cosine_map@100	0.1376

Training Details

Training Dataset

Unnamed Dataset

Size: 9,593 training samples
Columns: positive and anchor
Approximate statistics based on the first 1000 samples:
positive anchor
type string string
details
min: 3 tokens
mean: 49.72 tokens
max: 190 tokens
min: 9 tokens
mean: 21.22 tokens
max: 45 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.`	`Quin és el paper de l'Ajuntament en la comunicació de modificació d'activitat?`
`El Carnet Blau és un carnet personal i intransferible que acredita el compliment dels requisits per a gaudir d'un conjunt de descomptes i avantatges.`	`Quin és el propòsit del Carnet Blau en relació amb els descomptes?`
`Bonificació del 25% de l'import corresponent al consum d'aigua, la conservació d'escomeses, aforaments i comptadors així com els drets de connexió.`	`Quin és l'objectiu de la bonificació de la taxa per distribució i subministrament d'aigua?`

Loss: MatryoshkaLoss with these parameters:
json
1{ 2 "loss": "MultipleNegativesRankingLoss", 3 "matryoshka_dims": [ 4 1024, 5 768, 6 512, 7 256, 8 128, 9 64 10 ], 11 "matryoshka_weights": [ 12 1, 13 1, 14 1, 15 1, 16 1, 17 1 18 ], 19 "n_dims_per_step": -1 20}

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
0.2667	10	3.4587	-	-	-	-	-	-
0.5333	20	2.8693	-	-	-	-	-	-
0.8	30	2.3094	-	-	-	-	-	-
0.9867	37	-	0.1331	0.1252	0.1322	0.1337	0.1128	0.1347
1.0667	40	1.6196	-	-	-	-	-	-
1.3333	50	1.1926	-	-	-	-	-	-
1.6	60	0.9497	-	-	-	-	-	-
1.8667	70	0.882	-	-	-	-	-	-
2.0	75	-	0.1372	0.1272	0.1298	0.1365	0.1212	0.1369
2.1333	80	0.5621	-	-	-	-	-	-
2.4	90	0.4454	-	-	-	-	-	-
2.6667	100	0.4143	-	-	-	-	-	-
2.9333	110	0.4014	-	-	-	-	-	-
2.9867	112	-	0.1365	0.1282	0.1329	0.1437	0.1259	0.1390
3.2	120	0.2863	-	-	-	-	-	-
3.4667	130	0.1977	-	-	-	-	-	-
3.7333	140	0.2411	-	-	-	-	-	-
4.0	150	0.222	0.1355	0.1308	0.1378	0.1346	0.1239	0.1362
4.2667	160	0.1705	-	-	-	-	-	-
4.5333	170	0.1522	-	-	-	-	-	-
4.8	180	0.1606	-	-	-	-	-	-
4.9867	187	-	0.1441	0.1305	0.1344	0.1373	0.1356	0.1409
5.0667	190	0.1281	-	-	-	-	-	-
5.3333	200	0.1099	-	-	-	-	-	-
5.6	210	0.0921	-	-	-	-	-	-
5.8667	220	0.114	-	-	-	-	-	-
6.0	225	-	0.1371	0.1361	0.1331	0.1371	0.1351	0.1421
6.1333	230	0.0703	-	-	-	-	-	-
6.4	240	0.0746	-	-	-	-	-	-
6.6667	250	0.0734	-	-	-	-	-	-
6.9333	260	0.0803	-	-	-	-	-	-
6.9867	262	-	0.1447	0.1400	0.1422	0.1397	0.1376	0.1395
7.2	270	0.0684	-	-	-	-	-	-
7.4667	280	0.0493	-	-	-	-	-	-
7.7333	290	0.0531	-	-	-	-	-	-
8.0	300	0.0705	0.1410	0.1411	0.1379	0.1372	0.1372	0.1380
8.2667	310	0.0495	-	-	-	-	-	-
8.5333	320	0.0478	-	-	-	-	-	-
8.8	330	0.0455	-	-	-	-	-	-
8.9867	337	-	0.1463	0.1456	0.1416	0.1445	0.1408	0.1427
9.0667	340	0.0495	-	-	-	-	-	-
9.3333	350	0.0457	-	-	-	-	-	-
9.6	360	0.0487	-	-	-	-	-	-
9.8667	370	0.0568	0.1464	0.1416	0.1415	0.1464	0.1376	0.1454

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

bibtex

@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

bibtex

@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

bibtex

@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}
}

	positive	anchor
type	string	string
details	min: 3 tokens mean: 49.72 tokens max: 190 tokens	min: 9 tokens mean: 21.22 tokens max: 45 tokens