metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:6000
  - loss:CoSENTLoss
base_model: avsolatorio/GIST-small-Embedding-v0
widget:
  - source_sentence: are paris metro tickets one way?
    sentences:
      - >-
        The two big differences between the 2.4 GHz and 5 GHz frequencies are
        speed and range. A wireless transmission at 2.4 GHz provides internet to
        a larger area but sacrifices speed, while 5 GHz provides faster speeds
        to a smaller area.
      - >-
        The State of Rhode Island has adopted the income shares model to
        determine the weekly child support order. It is based upon the
        philosophy that children are entitled to the standard of living based
        upon both parents monthly income. ... Weekly gross income of both
        parents before taxes and before any other deductions.
      - >-
        Insulin NPH may be administered in 2 divided doses daily (either as
        equally divided doses, or as ~2/3 of the dose before the morning meal
        and ~1/3 of the dose before the evening meal or at bedtime).
  - source_sentence: how to pxe boot surface pro?
    sentences:
      - >-
        The UKTV Play app, with shows from Dave, Drama, Yesterday and Really, is
        available on smart TVs powered by Freeview Play and newer Samsung TVs.
        ... You can watch catch up and box sets from W, Alibi, Gold, Eden, Dave,
        Drama and Yesterday on Sky+HD, Sky Q and Sky Go.
      - >-
        In a branch For cash that was deposited over the counter at another
        bank, the processing and clearance time is 5 business days (not
        including public holidays).
      - >-
        ['Click "account" in the upper right corner of your Facebook page.',
        'Select "privacy settings."', 'Under "block lists" at the bottom center
        of the page, click "edit your lists."', 'At the top, under "block
        users," add the name or e-mail address of the person you\'d like to
        block.', 'Click "block."']
  - source_sentence: what is long-term capital gains rate?
    sentences:
      - >-
        You can get Social Security retirement or survivors benefits and work at
        the same time. But, if you're younger than full retirement age, and earn
        more than certain amounts, your benefits will be reduced. The amount
        that your benefits are reduced, however, isn't truly lost.
      - >-
        Dreams that involve shouting can warn of impending trouble. When you are
        the one shouting, this can mean you are going through a tough time in
        your waking life. You may be only feeling only negative emotions. ...
        Hearing someone else shouting signifies a warning of fright or anger.
      - >-
        A regular polygon is a flat shape whose sides are all equal and whose
        angles are all equal. The formula for finding the sum of the measure of
        the interior angles is (n - 2) * 180. To find the measure of one
        interior angle, we take that formula and divide by the number of sides
        n: (n - 2) * 180 / n.
  - source_sentence: can a girl get pregnant two days after her menstruation?
    sentences:
      - >-
        Newborn usually refers to a baby from birth to about 2 months of age.
        Infants can be considered children anywhere from birth to 1 year old.
        Baby can be used to refer to any child from birth to age 4 years old,
        thus encompassing newborns, infants, and toddlers.
      - >-
        According to professional numerologists, there are three ultimately
        lucky numbers for Capricorn-born people: they are 5, 6, and 8. In case
        they want to increase the chance of success for anything, simply make
        use of these numbers.
      - >-
        He's a professional dancer and model. J.C. Before entering the Big
        Brother house, J.C. was a dancer who traveled the world to perform
        professionally. “I do professional dancing. Not really break dancing, I
        do more choreography dancing,” he said in an interview with
        Entertainment Tonight Canada.
  - source_sentence: how long does it take to transfer money between anz and westpac?
    sentences:
      - >-
        This service is currently offered free of charge by the bank. You can
        get the last 'Available' balance of your account (by an SMS) by giving a
        Missed Call to 18008431122. You can get the Mini Statement (by an SMS)
        for last 5 transactions in your account by giving a Missed Call to
        18008431133. 1.
      - >-
        Simply put, 1 ply toilet paper is made of a single layer of paper, while
        2 ply has two layers. ... In the 1950's, a manufacturer created a method
        to roll and attach one-ply paper together to make a thicker “two-ply”.
        For years, 2-ply toilet tissue was always thicker and usually assumed to
        be better.
      - >-
        The main difference between unique and distinct is that UNIQUE is a
        constraint that is used on the input of data and ensures data integrity.
        While DISTINCT keyword is used when we want to query our results or in
        other words, output the data.
pipeline_tag: sentence-similarity
library_name: sentence-transformers

SentenceTransformer based on avsolatorio/GIST-small-Embedding-v0

This is a sentence-transformers model finetuned from avsolatorio/GIST-small-Embedding-v0. 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: avsolatorio/GIST-small-Embedding-v0
Maximum Sequence Length: 512 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': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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("moshew/gist_small_ft_gooaq_v1")
# Run inference
sentences = [
    'how long does it take to transfer money between anz and westpac?',
    "This service is currently offered free of charge by the bank. You can get the last 'Available' balance of your account (by an SMS) by giving a Missed Call to 18008431122. You can get the Mini Statement (by an SMS) for last 5 transactions in your account by giving a Missed Call to 18008431133. 1.",
    "Simply put, 1 ply toilet paper is made of a single layer of paper, while 2 ply has two layers. ... In the 1950's, a manufacturer created a method to roll and attach one-ply paper together to make a thicker “two-ply”. For years, 2-ply toilet tissue was always thicker and usually assumed to be better.",
]
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: 6,000 training samples
Columns: sentence1, sentence2, and label
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 label
type string string float
details
min: 8 tokens
mean: 11.97 tokens
max: 23 tokens

min: 14 tokens
mean: 58.86 tokens
max: 126 tokens

min: 0.0
mean: 0.17
max: 1.0

	sentence1	sentence2	label
type	string	string	float
details	min: 8 tokens mean: 11.97 tokens max: 23 tokens	min: 14 tokens mean: 58.86 tokens max: 126 tokens	min: 0.0 mean: 0.17 max: 1.0

Samples:

sentence1	sentence2	label
`what is the difference between rapid rise yeast and bread machine yeast?`	`Though there are some minor differences in shape and nutrients, Rapid-Rise Yeast is (pretty much) the same as Instant Yeast and Bread Machine Yeast. ... Also, Rapid-Rise Yeast is a little more potent than Active Dry Yeast and can be mixed in with your dry ingredients directly.`	`1.0`
`what is the difference between rapid rise yeast and bread machine yeast?`	`Omeprazole and esomeprazole therapy are both associated with a low rate of transient and asymptomatic serum aminotransferase elevations and are rare causes of clinically apparent liver injury.`	`0.0`
`what is the difference between rapid rise yeast and bread machine yeast?`	`Benefits of choosing a soft starter A variable frequency drive (VFD) is a motor control device that protects and controls the speed of an AC induction motor. A VFD can control the speed of the motor during the start and stop cycle, as well as throughout the run cycle.`	`0.0`

Loss: CoSENTLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 1
warmup_ratio: 0.1
seed: 12
bf16: True
dataloader_num_workers: 4

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: 1.0
num_train_epochs: 1
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.1
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: 12
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: True
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: 4
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}
tp_size: 0
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: proportional

Training Logs

Epoch	Step	Training Loss
0.0027	1	0.3104

Framework Versions

Python: 3.11.12
Sentence Transformers: 4.1.0
Transformers: 4.51.3
PyTorch: 2.6.0+cu124
Accelerate: 1.5.2
Datasets: 3.5.0
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",
}

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},
}