metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - generated_from_trainer
  - dataset_size:24593
  - loss:CoSENTLoss
base_model: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
  - pearson_manhattan
  - spearman_manhattan
  - pearson_euclidean
  - spearman_euclidean
  - pearson_dot
  - spearman_dot
  - pearson_max
  - spearman_max
model-index:
  - name: >-
      SentenceTransformer based on
      sentence-transformers/finetuned_paraphrase-multilingual-MiniLM-L12-v2
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: pearson_cosine
            value: 0.03594393239556079
            name: Pearson Cosine
          - type: spearman_cosine
            value: -0.00047007527052389596
            name: Spearman Cosine
          - type: pearson_manhattan
            value: 0.02486157492330912
            name: Pearson Manhattan
          - type: spearman_manhattan
            value: -0.002126248151952068
            name: Spearman Manhattan
          - type: pearson_euclidean
            value: 0.024692776461385596
            name: Pearson Euclidean
          - type: spearman_euclidean
            value: -0.0020342683424227027
            name: Spearman Euclidean
          - type: pearson_dot
            value: -0.005055107350691934
            name: Pearson Dot
          - type: spearman_dot
            value: 0.0015424580293819054
            name: Spearman Dot
          - type: pearson_max
            value: 0.03594393239556079
            name: Pearson Max
          - type: spearman_max
            value: 0.0015424580293819054
            name: Spearman Max
license: mit
language:
  - nl

SentenceTransformer based on sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for stylistic and semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more. I personally used this to give LLM generated sentences a rating between 0 and 1 on how good they match the style of the city of Antwerp.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
Maximum Sequence Length: 128 tokens
Output Dimensionality: 384 tokens
Similarity Function: Cosine Similarity
Language: Dutch, Flemish

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: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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("sentence_transformers_model_id")
# Run inference
sentences = [
    '"Daarnaast willen ze hun bestaande platform DETECT, waarmee onderzoekers unieke inzichten kunnen verwerven in de respons tegen een vaccin, commercialiseren."',
    '"Ze zijn van plan om het platform DETECT, dat onderzoekers helpt bij het verkrijgen van unieke inzichten over hoe een vaccin reageert, verder te ontwikkelen en commercieel beschikbaar te maken."',
    '"In februari 2020 hield buurtcomit Stadspark een eerste gesprek over het Stadspark."',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Semantic Similarity

Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.0359
spearman_cosine	-0.0005
pearson_manhattan	0.0249
spearman_manhattan	-0.0021
pearson_euclidean	0.0247
spearman_euclidean	-0.002
pearson_dot	-0.0051
spearman_dot	0.0015
pearson_max	0.0359
spearman_max	0.0015

Training Details

Training Dataset

Unnamed Dataset

Size: 24,593 training samples
Columns: sentence1, sentence2, and label
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 label
type string string float
details
min: 18 tokens
mean: 34.72 tokens
max: 128 tokens

min: 10 tokens
mean: 34.48 tokens
max: 128 tokens

min: 0.0
mean: 0.63
max: 1.0

	sentence1	sentence2	label
type	string	string	float
details	min: 18 tokens mean: 34.72 tokens max: 128 tokens	min: 10 tokens mean: 34.48 tokens max: 128 tokens	min: 0.0 mean: 0.63 max: 1.0

Samples:

sentence1	sentence2	label
`"Bij een noodsituatie zoals een grote brand, een overstroming of een stroomonderbreking stuurt BE-Alert automatisch berichten uit."`	`"In een noodgeval zoals een grote brand, een overstroming of een stroomuitval, waarschuwt BE-Alert ons direct via sms."`	`1.0`
`"Nationale test BE-Alert 18 steden en gemeenten in de provincie Antwerpen namen deel aan de nationale test op donderdag 7 oktober 2021."`	`"In de provincie Antwerpen deden 18 stadsdelen en districten mee aan de nationale test van BE-Alert op donderdag 7 oktober 2021."`	`0.9`
`"Vrouwen van 50 tot 69 jaar die de voorbije 2 jaar geen mammografie lieten maken, ontvangen een uitnodiging voor een gratis mammografie."`	`"Vrouwen tussen de 50 en 69 jaar die de afgelopen twee jaar geen mammografie hebben laten doen, ontvangen een uitnodiging voor een gratis mammografie."`	`1.0`

Loss: CoSENTLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "pairwise_cos_sim"
}

Evaluation Dataset

Unnamed Dataset

Size: 10,540 evaluation samples
Columns: sentence1, sentence2, and label
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 label
type string string float
details
min: 18 tokens
mean: 37.23 tokens
max: 128 tokens

min: 13 tokens
mean: 36.14 tokens
max: 128 tokens

min: 0.0
mean: 0.64
max: 1.0

	sentence1	sentence2	label
type	string	string	float
details	min: 18 tokens mean: 37.23 tokens max: 128 tokens	min: 13 tokens mean: 36.14 tokens max: 128 tokens	min: 0.0 mean: 0.64 max: 1.0

Samples:

sentence1	sentence2	label
`"Op dinsdag 23 mei verschijnt de Stadskroniek ‘Tingeling. 150 jaar tram in Antwerpen’ Deze Stadskroniek neemt de lezer mee in het dagelijkse leven van de reizigers en de bemanning van de trams in Antwerpen."`	`"Op dinsdag 23 mei verschijnt de Stadskroniek 'Tingeling. 150 jaar tram in Antwerpen'. Deze Stadskroniek neemt je mee in het dagelijkse leven van de reizigers en de bemanning van de trams in Antwerpen."`	`1.0`
"De pers wordt vriendelijk uitgenodigd op de lancering van de Stadskroniek ‘Tingeling. 150 jaar tram in Antwerpen’ op dinsdag 23 mei om 20 uur in het Vlaams Tram- en Autobusmuseum, Diksmuidelaan 42, 2600 Antwerpen Verwelkoming door Bob Morren, auteur Toespraak door Nabilla Ait Daoud, schepen voor cultuur Toespraak door Koen Kennis, schepen voor mobiliteit Korte gegidste rondleiding in het trammuseum door Bob Morren Stadskronieken zijn erfgoedverhalen over Antwerpen en de Antwerpse districten."	"De pers is van harte uitgenodigd voor de lancering van 'Tingeling. 150 jaar tram in Antwerpen' op dinsdag 23 mei om 20 uur bij het Vlaams Tram- en Autobusmuseum, Diksmuidelaan 42, in Antwerpen. Bob Morren, bekend van zijn boek 'Toespraak door Nabilla Ait Daoud, schepen voor cultuur, zal de avond openen met een welkomstwoord. Ook Koen Kennis, schepen voor mobiliteit, spreekt over de impact van trams op onze stad. Na deze toespraken volgt een korte rondleiding door Bob Morren in het museum. Stadskronieken zijn verhalen die ons erfgoed vieren en leren over Antwerpen en haar districten."	`1.0`

                                                                                                                                                                                                                                      | <code>0.9</code> |

Loss: CoSENTLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "pairwise_cos_sim"
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
learning_rate: 4e-06
num_train_epochs: 2
fp16: True
load_best_model_at_end: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 4e-06
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 2
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
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: False
fp16: True
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
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
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
use_liger_kernel: False
eval_use_gather_object: False
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	spearman_cosine
0.1664	128	-	5.8279	-0.0016
0.3329	256	-	5.8067	-0.0052
0.4993	384	-	5.8030	-0.0042
0.6502	500	5.997	-	-
0.6658	512	-	5.8018	-0.0036
0.8322	640	-	5.8020	-0.0023
0.9987	768	-	5.8033	-0.0021
1.1651	896	-	5.8056	-0.0012
1.3004	1000	5.7987	-	-
1.3316	1024	-	5.8079	-0.0017
1.4980	1152	-	5.8090	-0.0015
1.6645	1280	-	5.8033	-0.0005
1.8309	1408	-	5.8039	-0.0003
1.9506	1500	5.8021	-	-
1.9974	1536	-	5.8043	-0.0005

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.11.10
Sentence Transformers: 3.2.0
Transformers: 4.45.0
PyTorch: 2.5.1+cu124
Accelerate: 1.1.1
Datasets: 3.1.0
Tokenizers: 0.20.3

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",
}

CoSENTLoss

@online{kexuefm-8847,
    title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
    author={Su Jianlin},
    year={2022},
    month={Jan},
    url={https://kexue.fm/archives/8847},
}