SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. It maps sentences & paragraphs to a 384-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-MiniLM-L6-v2
Maximum Sequence Length: 256 tokens
Output Dimensionality: 384 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': 256, '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})
  (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 = [
    'Art is sometimes divided into two kinds, high art and popular art. High art attracts a much smaller population than popular art,but the number is large and growing. People who enjoy high art go to the opera and symphony concerts ; they read serious books and go to serious plays ; they keep up with art exhibitions. Popular art is mainly a kind of amusement. Some TV programs are meant to be watched today and forgotten tomorrow. Many popular songs are hits for a few weeks ; then they disappear. Other songs remain popular for such a long time that they become _ . The line between high and popular art is not always clear,however. Many people believe that rock music, for example, is a real art form. Many films are also taken seriously ,while others disappear as nothing more than amusement. Many more people like popular art than high art because   _  . A. popular art is better than high art. B. high art is not a real art form. C. popular art will be loved for a longer time than high art. D. popular art is a kind of amusement.',
    '**The arts**\n\nThe arts:\nThe arts are a very wide range of human practices of creative expression, storytelling and cultural participation. They encompass multiple diverse and plural modes of thinking, doing and being, in an extremely broad range of media. Both highly dynamic and a characteristically constant feature of human life, they have developed into innovative, stylized and sometimes intricate forms. This is often achieved through sustained and deliberate study, training and/or theorizing within a particular tradition, across generations and even between civilizations. The arts are a vehicle through which human beings cultivate distinct social, cultural and individual identities, while transmitting values, impressions, judgments, ideas, visions, spiritual meanings, patterns of life and experiences across time and space.',
    '**Genre painting**\n\nGenre painting:\nGenre painting (or petit genre), a form of genre art, depicts aspects of everyday life by portraying ordinary people engaged in common activities. One common definition of a genre scene is that it shows figures to whom no identity can be attached either individually or collectively, thus distinguishing it from history paintings (also called grand genre) and portraits. A work would often be considered as a genre work even if it could be shown that the artist had used a known person—a member of his family, say—as a model. In this case it would depend on whether the work was likely to have been intended by the artist to be perceived as a portrait—sometimes a subjective question. The depictions can be realistic, imagined, or romanticized by the artist. Because of their familiar and frequently sentimental subject matter, genre paintings have often proven popular with the bourgeoisie, or middle class.',
]
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]

Training Details

Training Dataset

Unnamed Dataset

Size: 117,937 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: 25 tokens mean: 121.28 tokens max: 256 tokens	min: 10 tokens mean: 141.37 tokens max: 256 tokens	min: 10 tokens mean: 140.67 tokens max: 256 tokens

Samples:

sentence_0	sentence_1	sentence_2
`What organ do we use to hear sound? A. eye. B. the ear. C. amplifier. D. antennae.`	`The organ that we use to hear sound is the ear. Almost all the structures in the ear are needed for this purpose. Together, they gather and amplify sound waves and change their energy to electrical signals. The electrical signals travel to the brain, which interprets them as sound.`	`Mouth organ Mouth organ: A mouth organ is any free reed aerophone with one or more air chambers fitted with a free reed. Though it spans many traditions, it is played universally the same way by the musician placing their lips over a chamber or holes in the instrument, and blowing or sucking air to create a sound. Many of the chambers can be played together or each individually.`
`Preveebral space thickness in adult at C6-C7 level is A. 7mm. B. 15mm. C. 22mm. D. 30mm.`	`Infraorbital margin Infraorbital margin: The infraorbital margin is the lower margin of the eye socket.`	Cervical enlargement Cervical enlargement: The cervical enlargement corresponds with the attachments of the large nerves which supply the upper limbs. Located just above the brachial plexus, it extends from about the fifth cervical to the first thoracic vertebra, its maximum circumference (about 38 mm.) being on a level with the attachment of the sixth pair of cervical nerves. The reason behind the enlargement of the cervical region is because of the increased neural input and output to the upper limbs. An analogous region in the lower limbs occurs at the lumbar enlargement.
`Grey Turner's sign (Flank discolouration) is seen in A. Acute pyelonephritis. B. Acute cholecystitis. C. Acute pancreatitis. D. Acute peritonitis.`	Bancroft's sign Bancroft's sign: Bancroft's sign, also known as Moses' sign, is a clinical sign found in patients with deep vein thrombosis of the lower leg involving the posterior tibial veins. The sign is positive if pain is elicited when the calf muscle is compressed forwards against the tibia, but not when the calf muscle is compressed from side to side. Like other clinical signs for deep vein thrombosis, such as Homans sign and Lowenberg's sign, this sign is neither sensitive nor specific for the presence of thrombosis.	`Dance's sign Dance's sign: Dance's sign is an eponymous medical sign consisting of an investigation of the right lower quadrant of the abdomen for retraction, which can be an indication of intussusception, i.e. those with intussusception may have retraction of the right iliac fossa.`

Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 1
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: 16
per_device_eval_batch_size: 16
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: 1
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
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: round_robin

Training Logs

Epoch	Step	Training Loss
0.0678	500	0.6209
0.1356	1000	0.6208
0.2035	1500	0.5913
0.2713	2000	0.5989
0.3391	2500	0.5879
0.4069	3000	0.584
0.4748	3500	0.5627
0.5426	4000	0.5564
0.6104	4500	0.5511
0.6782	5000	0.5511
0.7461	5500	0.5345
0.8139	6000	0.5334
0.8817	6500	0.5377
0.9495	7000	0.5227

Framework Versions

Python: 3.11.9
Sentence Transformers: 4.1.0
Transformers: 4.52.4
PyTorch: 2.7.1
Accelerate: 1.7.0
Datasets: 3.6.0
Tokenizers: 0.21.1

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",
}

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}
}