This is a sentence-transformers model trained. 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

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, 'architecture': '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'electronic noise rock',
    "Dead Combo. by Dead Combo. Genre: Rock; Released: 2004. Review: still applicable live precise good its used bloody loud brisk another dead sonic dynamic even stiff other unusual rhythmic most all baby two melodic sunny affable heavy vocal right potent such trying little shrill twisted hot stock noisy acoustic electronic Trevor Jackson's Output label offers the debut record from this Manhattan (by way of Finland) duo that's well-versed in both Jesus & Mary Chain jangle and stiff, propulsive rhythms.",
    'IV. by Faust. Genre: Rock; Released: 2006. Review: willing huge previous young just calculating either work slick hard still major ridiculous early tripping good pretty more seven long second its better fragmented outside happy classic professional alleged completed which threatened due sonic even seeming bad other third fourth exploited minute status smashing all super important sad relieved angelic new milky left messy jumbled impossible whose inspired electronic random disturbing half cut each palatable nan',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.3686, 0.2751],
#         [0.3686, 1.0000, 0.9454],
#         [0.2751, 0.9454, 1.0000]])

Training Details

Training Dataset

Unnamed Dataset

Size: 9,362 training samples
Columns: sentence_0, sentence_1, and sentence_2
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1 sentence_2
type string string string
details
min: 4 tokens
mean: 9.45 tokens
max: 25 tokens
min: 75 tokens
mean: 131.44 tokens
max: 253 tokens
min: 65 tokens
mean: 130.26 tokens
max: 215 tokens

Samples:

sentence_0	sentence_1	sentence_2
`house gospel country meditation`	Angels of Death. by Jennifer Castle. Genre: Folk/Country; Released: 2018. Review: clingy specific just work hard time home tragic long second burning poignant bouncing final rough steel rural electric breezy big conflicting instrumental resilient magical which universal divided relatable meandering inarticulate difficult sparse simple creative devastating personal two natural house similar heavy crying absurd first charming flooded trying driven down rude full least fitting inspired glass acoustic her each The Toronto singer’s album is a country and gospel-infused meditation on death and mourning that flickers between the broadly universal and the devastatingly personal.	`Greatest Palace Music. by Bonnie “Prince” Billy. Genre: Folk/Country; Released: 2004. Review: bizarre human favorite young just mocking work legendary hallowed following its long his biggest haggard learned lofty chance crazy even bad soulful personal black house splattered informed perplexed first such smooth one fresh selected these greatest novel present different The first country record I ever bought was Merle Haggard's 16 Biggest Hits. Splattered into a harrowing dustbowl of ...`
`disco country ethereal rock pop`	Walking on a Dream. by Empire of the Sun. Genre: Electronic / Rock; Released: 2008. Review: impenetrable still just ridiculous either immediate indulgent divergent its relentless following sublime second cutting psychedelic his same unlikely odd old big ethereal chance simplistic uneasy complex material cleared which much best even cut snappy surprised high diverging grand both kind worst two middle sunny straight puzzling right first such poised new ready down little individual real overwrought meaningless full incapable these acid last late historical fitting white thematic inspired sleepy tough half trenchant each The first full-length collaboration from Australians Luke Steele of the Sleepy Jackson and Nick Littlemore of Pnau works on a grand scale as it moves between streamlined psychedelia and flamboyant dance-pop.	Morning Phase. by Beck. Genre: Rock; Released: 2014. Review: favorite previous ambiguous time neutral live early good stunning more vague long playful its used following involved flat his same old ethereal which only another much alien excellent due sonic even dull alive entire comfortable straightforward overall minute less most lost two all red similar safe straight modern forthright queasy first new great working down unabashed daring heartbroken interesting one frustrating married these last fine sea indirect glass surprising On Morning Phase, Beck returns to the style and form of his 2002 singer-songwriter record Sea Change. Instead of a much-needed daring comeback statement, it feels more like a pointed exhale.
`folk warm drama pop saxophone`	Fear of Flying. by David Karsten Daniels. Genre: Folk/Country; Released: 2008. Review: still time live more basic long second its available scraggly born final his royal physical biggest familiar old fifth single musical away complex sharp former well only light universal which rustic compelling even cut female nice powerful halting devastating flying warm kind slow black all most blasphemous life every baby limited real trying left vigorous down lyrical modest one unfettered raging catastrophic particular inspired weary north otherwise David Karsten Daniels' latest is a collection of songs inspired by his grandparents' physical decline, songs that flip thoughtfully through some complex, if well-thumbed, existential issues: mortality, God, and the struggle for faith.	Stay Gold. by First Aid Kit. Genre: Folk/Country; Released: 2014. Review: dusty still day work unromantic just hard inviting gold home literary good wide more its outside motivational his winning affectionate popular undeniable big musical ideal mission evening even high other third soaring most choice both kind signed two natural all warm sunny first consistent new down mature these least male stoic weary convinced her each green Swedish folk-pop duo First Aid Kit's third LP and first for Columbia is their grandest and arguably their most consistent release to date, featuring weary songs about transience that approach their conflicts with wiseness and maturity.

Loss: TripletLoss with these parameters:
json
1{ 2 "distance_metric": "cosine_distance_func", 3 "triplet_margin": 0.2 4}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 32
per_device_eval_batch_size: 32
num_train_epochs: 2
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
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: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1
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
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}
parallelism_config: 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
project: huggingface
trackio_space_id: trackio
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
hub_revision: None
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
include_tokens_per_second: False
include_num_input_tokens_seen: no
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: True
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
1.7065	500	0.137

Framework Versions

Python: 3.12.12
Sentence Transformers: 5.1.2
Transformers: 4.57.2
PyTorch: 2.9.0+cu126
Accelerate: 1.12.0
Datasets: 4.0.0
Tokenizers: 0.22.1

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",
}

TripletLoss