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 = [
    'If the difference between the length and breadth of a rectangle is 23 m and its perimeter is 206 m, what is its area? A. 2510. B. 2530. C. 2515. D. 2520.',
    "A circle with radius $3$ has a sector with a $345^\\circ$ central angle. What is the area of the sector? ${9\\pi}$ $\\color{#9D38BD}{345^\\circ}$ ${\\dfrac{69}{8}\\pi}$ ${3}$\n\nHints:\nFirst, calculate the area of the whole circle. Then the area of the sector is some fraction of the whole circle's area. $A_c = \\pi r^2$ $A_c = \\pi (3)^2$ $A_c = 9\\pi$ The ratio between the sector's central angle $\\theta$ and $360^\\circ$ is equal to the ratio between the sector's area, $A_s$ , and the whole circle's area, $A_c$ $\\dfrac{\\theta}{360^\\circ} = \\dfrac{A_s}{A_c}$ $\\dfrac{345^\\circ}{360^\\circ} = \\dfrac{A_s}{9\\pi}$ $\\dfrac{23}{24} = \\dfrac{A_s}{9\\pi}$ $\\dfrac{23}{24} \\times 9\\pi = A_s$ $\\dfrac{69}{8}\\pi = A_s$",
    'The area of a rectangular surface is calculated as its length multiplied by its width.',
]
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: 115,928 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: 172.44 tokens max: 256 tokens	min: 10 tokens mean: 128.32 tokens max: 256 tokens	min: 10 tokens mean: 116.61 tokens max: 256 tokens

Samples:

sentence_0	sentence_1	sentence_2
Back in time, dog sledding was the only method of transportation in frozen parts of the world. You, too, can participate in this time-honored tradition after looking into dog sledding adventures and tours. Alaska Dog Sledding It offers two dog sledding tours in Alaska, the Golsovia Dog Trip and the Iditarod Dog Trip. On both trips, you will learn how to drive your own dog team. The Golsovia Dog Trip lasts six days. The Iditarod Dog Trip is held just before the internationally famous Iditarod race so that participants will be able to attend the Iditarod events. For more information, please see: Alaska Dog Sledding Greenland Expedition Specialists Travel through the shining ice and snow on the land of East Greenland via dog sledding. During this dog sledding vacation, you will be camping during the cold of winter and will take part in caring for the dogs and the camps. They also organize and guide kiting trips, sea kayaking trips and mountaineering vacations. For more informa...	`Dog camp Dog camp: Dog camp is a form of vacation for owners accompanied by their dogs with dog-centric activities ranging from casual recreational playtime to serious obedience or sport training. In many dog camps dogs can play and socialize throughout the day while supervised by their owners. Some of the activities at a dog camp might include running, fetching balls or frisbees, chasing other dogs, tug of war, socializing amongst playmates, and lessons.`	`Dog camp Dog camp: Dog camp is a form of vacation for owners accompanied by their dogs with dog-centric activities ranging from casual recreational playtime to serious obedience or sport training. In many dog camps dogs can play and socialize throughout the day while supervised by their owners. Some of`
`My brother is 3 years elder to me.My father was 28 years of age when my sister was born while my mother was 26 years of age when I was born.If my sister was 4 years of age when my brother was born,then,what was the age of my father and mother respectively when my brother was born ? A. 32 yrs,23 yrs. B. 35 yrs,29 yrs. C. 35 yrs,33 yrs. D. None of these.`	`Abby is $3$ years old. Her brother Ben is $4$ years older than she is. How old is Ben? Hints: To find how old ${\text{Ben}}$ is, we start with how old ${\text{Abby}}$ is, and add ${4\text{ years}}$. $?$ $3$ $4$ Ben's age Abby's age 4 more years ${3} + {4} = {\Box}$ Let's add to find how old ${\text{Ben}}$ is. $$ ${+}$ $$ ${=}$ $$ $$ ${3} + {4} = {7}$ ${\text{Ben}}$ is ${7}$ years old.`	Ishaan is $3$ times as old as Christopher and is also $14$ years older than Christopher. How old is Ishaan? Hints: We can use the given information to write down two equations that describe the ages of Ishaan and Christopher. Let Ishaan's current age be $i$ and Christopher's current age be $c$. ${i = 3c}$ ${i = c + 14}$ Now we have two independent equations, and we can solve for our two unknowns. One way to solve for $i$ is to solve the second equation for $c$ and substitute that value into the first equation. Solving our second equation for $c$, we get: ${c = i - 14}$. Substituting this into our first equation, we get the equation: ${i = 3}{(i - 14)}$ which combines the information about $i$ from both of our original equations. Simplifying the right side of this equation, we get: $i = 3i - 42$. Solving for $i$, we get: $2 i = 42$. $i = 21$.
`Blood specimen for neonatal thyroid screening is obtained at: A. Cord blood. B. 24 hours after bih. C. 48 hours after bih. D. 72 hours after bih.`	`TRH stimulation test TRH stimulation test: Prior to the availability of sensitive TSH assays, thyrotropin releasing hormone or TRH stimulation tests were relied upon for confirming and assessing the degree of suppression in suspected hyperthyroidism. Typically, this stimulation test involves determining basal TSH levels and levels 15 to 30 minutes after an intravenous bolus of TRH. Normally, TSH would rise into the concentration range measurable with less sensitive TSH assays.`	Blood gas test Blood gas test: A blood gas test or blood gas analysis tests blood to measure blood gas tension values, it also measures blood pH, and the level and base excess of bicarbonate. The source of the blood is reflected in the name of each test; arterial blood gases come from arteries, venous blood gases come from veins and capillary blood gases come from capillaries. The blood gas tension levels of partial pressures can be used as indicators of ventilation, respiration and oxygenation. Analysis of paired arterial and venous specimens can give insights into the aetiology of acidosis in the newborn.

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
0.0690	500	0.6132
0.1380	1000	0.6121
0.2070	1500	0.5883
0.2760	2000	0.5838
0.3450	2500	0.5689
0.4140	3000	0.55
0.4830	3500	0.5422
0.5520	4000	0.5257
0.6210	4500	0.5099
0.6900	5000	0.5008
0.7590	5500	0.5066
0.8280	6000	0.4941
0.8970	6500	0.4881
0.9661	7000	0.4898

Framework Versions

Python: 3.12.7
Sentence Transformers: 4.1.0
Transformers: 4.52.3
PyTorch: 2.7.0
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}
}