metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:32351
  - loss:TripletLoss
base_model: sentence-transformers/all-mpnet-base-v2
widget:
  - source_sentence: >-
      Genetic conditions that cause nutritional deficiencies can prevent a
      person from removing meat from their diet.
    sentences:
      - >-
        Ante un estado que no quiere hablar del tema, para Cataluña, solo es
        posible seguir su propio camino por otras vías.
      - >-
        Retinol deficiency is a genetically pre-disposed condition that prevents
        conversion beta-carotene to Vitamin A \(retinol\) in humans. Since
        plants have no retinol \(only beta-carotene\), humans with this
        condition cannot have a vegan diet, only one with animal products.
      - >-
        People with hemochromatosis \(a genetic condition\) can benefit greatly
        from a vegan diet, due to the lower absorbing non-heme iron in plants
        \(compared to heme iron in meat\).
  - source_sentence: >-
      The definition of veganism is: "A way of living which seeks to exclude, as
      far as is possible and practicable, all forms of exploitation of, and
      cruelty to, animals for food, clothing or any other purpose." In the
      \(unlikely\) case of survival or health concerns, the "as far as possible
      and practicable" clause makes it possible for such persons to be
      considered vegan as they would have no alternative options.
    sentences:
      - >-
        Veganism is not solely about diet. A person can still choose to live in
        accordance with vegan values, such as by avoiding animal circuses and
        leather/fur products.
      - >-
        It's easier to regulate established companies in a legal market than it
        is in the black market. Any issue would be with bad regulations not
        legalization.
      - >-
        That definition is too vague. There are different definitions of
        veganism, many of which are not compatible with using animals in any
        circumstances. In a way we are all vegan depending on how easy you
        believe it is to reach all the necessary nutrition in your city harming
        as few animals as possible.
  - source_sentence: >-
      Adding coding to the school curriculum means that something else must be
      left out.
    sentences:
      - Coding skills are much needed in today's job market.
      - Cataluña saldría de la UE con efectos económicos desastrosos.
      - >-
        Teaching coding effectively is impossible unless teachers are trained
        appropriately first.
  - source_sentence: >-
      Animals have innate, individual rights, which are taken away when they are
      killed or made to suffer.
    sentences:
      - Animals have a desire to live.
      - >-
        Uno de los ejemplos más claros es la falta de inversión reiterada al
        Corredor Mediterráneo  \(Algeciras-Valencia-Barcelona-Francia\),
        prioritario para la UE y Catalunya, pero relegado a algo residual por el
        estado Español.
      - >-
        A vegan society would equate humans rights with animal rights, which
        would make society worse off overall.
  - source_sentence: >-
      The sorts of people likely to lash out against affirmative action policies
      probably already hold negative views towards racial minorities.
    sentences:
      - >-
        The Far Right movement sees the inequality affirmative action addresses
        not as a problem to be solved, but as an outcome to be desired.
      - >-
        There are plenty of people who hold a positive view towards racial
        minorities and still oppose affirmative action.
      - >-
        Research has shown that college degrees have less economic utility for
        people from low socio-economic backgrounds.
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy
model-index:
  - name: SentenceTransformer based on sentence-transformers/all-mpnet-base-v2
    results:
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: cosine_accuracy
            value: 0.9264069199562073
            name: Cosine Accuracy
          - type: cosine_accuracy
            value: 0.9161931872367859
            name: Cosine Accuracy

SentenceTransformer based on sentence-transformers/all-mpnet-base-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-mpnet-base-v2. 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: sentence-transformers/all-mpnet-base-v2
Maximum Sequence Length: 384 tokens
Output Dimensionality: 768 dimensions
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': 384, 'do_lower_case': False}) with Transformer model: MPNetModel 
  (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})
  (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("sentence_transformers_model_id")
# Run inference
sentences = [
    'The sorts of people likely to lash out against affirmative action policies probably already hold negative views towards racial minorities.',
    'The Far Right movement sees the inequality affirmative action addresses not as a problem to be solved, but as an outcome to be desired.',
    'There are plenty of people who hold a positive view towards racial minorities and still oppose affirmative action.',
]
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

Triplet

Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9264

Triplet

Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9162

Training Details

Training Dataset

Unnamed Dataset

Size: 32,351 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative
type	string	string	string
details	min: 6 tokens mean: 30.94 tokens max: 160 tokens	min: 6 tokens mean: 40.8 tokens max: 180 tokens	min: 6 tokens mean: 44.95 tokens max: 162 tokens

Samples:

anchor	positive	negative
`La soberanía y la decisión sobre la unidad de España residen en el conjunto de España.`	`Apostar por un proceso de secesión es ir en contra de la globalización, la corriente histórica que vivimos.`	`Los tratados internacionales (incluido el Tratado de La Unión Europea) no serían aplicables a Cataluña como estado independiente, por lo que su permanencia en Europa podría verse interrumpida.`
`La soberanía y la decisión sobre la unidad de España residen en el conjunto de España.`	`Para sentar un precedente en conflictos de autodeterminación en el mundo.`	`La independencia de Cataluña afectaría negativamente a la economía de España.`
`La soberanía y la decisión sobre la unidad de España residen en el conjunto de España.`	`Para terminar con el trato injusto que recibe Cataluña al ser parte de España.`	`Por definición, cualquier nacionalismo es malo ya que crea divisiones artificiales y es fuente de conflictos.`

Loss: TripletLoss with these parameters:

{
    "distance_metric": "TripletDistanceMetric.COSINE",
    "triplet_margin": 0.3
}

Training Hyperparameters

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 8
per_device_eval_batch_size: 8
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: 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: 3.0
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: False
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: False
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: None
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
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
average_tokens_across_devices: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	cosine_accuracy
0.1236	500	0.1872	-
0.2473	1000	0.1954	-
0.3709	1500	0.1854	-
0.4946	2000	0.1891	-
0.6182	2500	0.181	-
0.7418	3000	0.1794	-
0.8655	3500	0.1815	-
0.9891	4000	0.1736	-
1.1128	4500	0.1342	-
1.2364	5000	0.1297	-
1.3600	5500	0.1318	-
1.4837	6000	0.1255	-
1.6073	6500	0.128	-
1.7310	7000	0.1233	-
1.8546	7500	0.1221	-
1.9782	8000	0.1232	-
2.1019	8500	0.0841	-
2.2255	9000	0.0757	-
2.3492	9500	0.0764	-
2.4728	10000	0.0761	-
2.5964	10500	0.0726	-
2.7201	11000	0.0644	-
2.8437	11500	0.073	-
2.9674	12000	0.0725	-
-1	-1	-	0.9162

Framework Versions

Python: 3.11.11
Sentence Transformers: 3.4.1
Transformers: 4.48.3
PyTorch: 2.5.1+cu124
Accelerate: 1.3.0
Datasets: 3.3.2
Tokenizers: 0.21.0

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification},
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}