DeBERTa2-0.9B-ST-v2

SentenceTransformer based on microsoft/deberta-v2-xlarge

This is a sentence-transformers model finetuned from microsoft/deberta-v2-xlarge on the negation-triplets, vitaminc-pairs, scitail-pairs-qa, scitail-pairs-pos, xsum-pairs, sciq_pairs, qasc_pairs, openbookqa_pairs, msmarco_pairs, nq_pairs, trivia_pairs, gooaq_pairs, paws-pos and global_dataset datasets. It maps sentences & paragraphs to a 1536-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: microsoft/deberta-v2-xlarge <!-- at revision 1d134961d4db8e7e8eb1bc1ab81cb370244c57f7 -->
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1536 tokens
• Similarity Function: Cosine Similarity
• Training Datasets:
  • negation-triplets
  • vitaminc-pairs
  • scitail-pairs-qa
  • scitail-pairs-pos
  • xsum-pairs
  • sciq_pairs
  • qasc_pairs
  • openbookqa_pairs
  • msmarco_pairs
  • nq_pairs
  • trivia_pairs
  • gooaq_pairs
  • paws-pos
  • global_dataset
• Language: en

<!-- - **License:** Unknown -->

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': False}) with Transformer model: DebertaV2Model 
  (1): Pooling({'word_embedding_dimension': 1536, '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})
)

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("bobox/DeBERTa2-0.9B-ST-v2")
# Run inference
sentences = [
    '"""Music of the Night"" and ""All I Ask of You"" are songs from which stage musical?"',
    'All I Ask of You - YouTube All I Ask of You Want to watch this again later? Sign in to add this video to a playlist. Need to report the video? Sign in to report inappropriate content. Rating is available when the video has been rented. This feature is not available right now. Please try again later. Published on Jan 6, 2012 Twenty-second in a series of clips from the 2004 film version of The Phantom of the Opera. Performed by Patrick Wilson as Raoul and Emmy Rossum as Christine. For more Phantom, find us on Facebook: http://www.facebook.com/ThePhantomOfT... Or follow us on Twitter: http://www.twitter.com/TheOperaGhosts Or add us to your Google+ Circles: https://plus.google.com/1019282273838... Category',
    'Beatles US Labels - Variations \xa0 THE CAPITOL � APPLE    LABEL DISCOGRAPHY\xa0 In early 1964, Capitol Records became    the primary U.S. manufacturer of The Beatles major releases.\xa0 When    Apple Records was formed by the Beatles in 1968, Capitol still maintained    production and distribution control over the Beatles records.\xa0 When    Apple was dissolved in 1975, the entire Beatles catalog, including solo    releases, reverted back to Capitol. In the early 1990�s, all The    Beatles/solo Capitol product was reissued on the Apple label.\xa0 The Beatles, collectively and    individually, have also appeared on several other labels over the    years.\xa0 However, most of these releases failed to stay in production    long enough to experience significant label design changes.\xa0 The    relatively few that did, are thoroughly identified in their respective    sections.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1536]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

<!-- ### Direct Usage (Transformers) <details><summary>Click to see the direct usage in Transformers</summary> </details> --> <!-- ### Downstream Usage (Sentence Transformers) You can finetune this model on your own dataset. <details><summary>Click to expand</summary> </details> --> <!-- ### Out-of-Scope Use *List how the model may foreseeably be misused and address what users ought not to do with the model.* -->

Evaluation

Metrics

Semantic Similarity

• Dataset: sts-test
• Evaluated with <code>EmbeddingSimilarityEvaluator</code>

Binary Classification

• Dataset: allNLI-dev
• Evaluated with <code>BinaryClassificationEvaluator</code>

Binary Classification

• Dataset: Qnli-dev
• Evaluated with <code>BinaryClassificationEvaluator</code>

<!-- ## Bias, Risks and Limitations *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.* --> <!-- ### Recommendations *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.* -->

Training Details

Training Datasets

negation-triplets

• Dataset: negation-triplets
• Size: 4,987 training samples
• Columns: <code>anchor</code>, <code>entailment</code>, and <code>negative</code>
• Approximate statistics based on the first 1000 samples:

  	anchor	entailment	negative
  type	string	string	string
  details	<ul><li>min: 4 tokens</li><li>mean: 21.87 tokens</li><li>max: 144 tokens</li></ul>	<ul><li>min: 5 tokens</li><li>mean: 13.72 tokens</li><li>max: 46 tokens</li></ul>	<ul><li>min: 4 tokens</li><li>mean: 13.94 tokens</li><li>max: 39 tokens</li></ul>

• Samples:

  anchor	entailment	negative
  <code>a very dirty toilet in a tiled bathroom</code>	<code>A dirty toilet in a dirty bathroom with a octagon tile floor.</code>	<code>A clean toilet in a dirty bathroom with an octagon tile floor.</code>
  <code>enjoy the wildlife</code>	<code>Enjoy the animals.</code>	<code>Ignore the animals.</code>
  <code>A man looking inside of birdcages on a sidewalk.</code>	<code>A man, holding his hat, is looking into the bird cage. </code>	<code>A man, holding his hat, is looking away from the bird cage. </code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

vitaminc-pairs

• Dataset: vitaminc-pairs at be6febb
• Size: 4,987 training samples
• Columns: <code>claim</code> and <code>evidence</code>
• Approximate statistics based on the first 1000 samples:

  	claim	evidence
  type	string	string
  details	<ul><li>min: 6 tokens</li><li>mean: 16.62 tokens</li><li>max: 54 tokens</li></ul>	<ul><li>min: 7 tokens</li><li>mean: 36.79 tokens</li><li>max: 133 tokens</li></ul>

• Samples:

  claim	evidence
  <code>Mase was born before 1976 .</code>	<code>Mase was born Mason Drell Betha in Jacksonville , Florida , on August 27 , 1975 , as a fraternal twin born almost two months premature .</code>
  <code>On Rotten Tomatoes , Going in Style received more than 105 reviews and a rating of under 47 % .</code>	<code>On Rotten Tomatoes , the film has an approval rating of 46 % based on 106 reviews , with an average rating of 5.3/10 .</code>
  <code>Aaron Charles Donald is an American football defensive end for Los Angeles Rams .</code>	<code>Aaron Charles Donald ( born May 23 , 1991 ) is an American football defensive end for the Los Angeles Rams of the National Football League ( NFL ) .</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

scitail-pairs-qa

• Dataset: scitail-pairs-qa at 0cc4353
• Size: 4,987 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 7 tokens</li><li>mean: 15.84 tokens</li><li>max: 41 tokens</li></ul>	<ul><li>min: 7 tokens</li><li>mean: 14.96 tokens</li><li>max: 41 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>A negative enthalpy change is observed in an exothermic reaction.</code>	<code>What enthalpy change is observed in an exothermic reaction?</code>
  <code>Fungus-like protists such as slime molds reproduce with spores.</code>	<code>How do fungus-like protists such as slime molds reproduce?</code>
  <code>Unlike energy, matter doesn’t need to be constantly added to ecosystems because it is recycled through ecosystems.</code>	<code>Unlike energy, what doesn’t need to be constantly added to ecosystems because it is recycled through ecosystems?</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

scitail-pairs-pos

• Dataset: scitail-pairs-pos at 0cc4353
• Size: 4,987 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 8 tokens</li><li>mean: 23.21 tokens</li><li>max: 64 tokens</li></ul>	<ul><li>min: 7 tokens</li><li>mean: 15.66 tokens</li><li>max: 39 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>During equinox times the Sun's vertical ray is shining directly on the Earth's equator and neither hemisphere is tilted toward or away from the Sun.</code>	<code>The sun is directly over the equator during.</code>
  <code>All the baby s major organs begin to develop in the first 6 to 8 weeks of pregnancy, so tight control from the moment of conception is critical.</code>	<code>By 8 weeks, all major organs start developing.</code>
  <code>Nobody disputes that all modern humans belong to one species, Homo sapiens .</code>	<code>Humans belong to the species homo sapiens.</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

xsum-pairs

• Dataset: xsum-pairs
• Size: 4,987 training samples
• Columns: <code>summary</code> and <code>document</code>
• Approximate statistics based on the first 1000 samples:

  	summary	document
  type	string	string
  details	<ul><li>min: 8 tokens</li><li>mean: 25.19 tokens</li><li>max: 44 tokens</li></ul>	<ul><li>min: 51 tokens</li><li>mean: 217.32 tokens</li><li>max: 422 tokens</li></ul>

• Samples:

  summary	document
  <code>Chester's Elliott Durrell salvaged a 1-1 draw at Boreham Wood with a late wonder strike.</code>	<code>Morgan Ferrier's fifth-minute effort looked like it would be enough for the hosts after the 24-year-old evaded Blaine Hudson to get in behind and tap the ball home.<br>After that the game became scrappy, with neither side creating many clear-cut chances, although Chester's Ryan Astles was forced into a goal-line clearance after 70 minutes to block a Matthew Paine drive.<br>Just when it looked like Chester boss Jon McCarthy was staring at a fourth defeat in his first five games, Durrell picked the ball up around halfway, spotted Grant Smith off his line and struck from distance to level things up with seven minutes remaining.<br>Report supplied by the Press Association<br>Match ends, Boreham Wood 1, Chester FC 1.<br>Second Half ends, Boreham Wood 1, Chester FC 1.<br>Substitution, Chester FC. Sam Hughes replaces Kane Richards.<br>Goal! Boreham Wood 1, Chester FC 1. Elliott Durrell (Chester FC).<br>Substitution, Boreham Wood. Jordan Chiedozie replaces Morgan Ferrier.<br>Substitution, Boreham Wood. Aaron Kuhl replaces Kenny Davis.<br>Substitution, Chester FC. Elliott Durrell replaces Jordan Chapell.<br>Evan Horwood (Chester FC) is shown the yellow card for a bad foul.<br>Second Half begins Boreham Wood 1, Chester FC 0.<br>First Half ends, Boreham Wood 1, Chester FC 0.<br>Goal! Boreham Wood 1, Chester FC 0. Morgan Ferrier (Boreham Wood).<br>First Half begins.<br>Lineups are announced and players are warming up.</code>
  <code>A major trauma centre in Stoke-on-Trent has been rated the best in the country for saving the lives of patients.</code>	<code>The University Hospitals of North Midlands Major Trauma Centre has the best total rolling survival rates of any adult major trauma single site centre since 2013.<br>Latest statistics show that for every 1,000 people treated in the last four years, 13 more survived than expected.<br>The centre is based at Royal Stoke University Hospital.<br>Medical director Dr John Oxtoby, described it as a "huge accomplishment".<br>See more stories from across Stoke and Staffordshire here<br>The figures come from the Trauma Audit and Research Network, an independent monitor of trauma care in England and Wales.<br>The data shows the centre also had the best survival rates for adult major trauma in 2015-16, when there were 15 extra survivors per 1,000 patients than expected.<br>"To have the best survival rates over four years of any major trauma centre is a phenomenal achievement," Dr Oxtoby added.<br>The centre treats patients from as far away as north Wales and the Peak District.<br>Those treated include people seriously injured in incidents such as vehicle crashes, falls, or assaults.</code>
  <code>Sunderland defender Paddy McNair will miss the rest of the season because of a cruciate knee ligament injury.</code>	<code>The 21-year-old, signed from Manchester United in August, has made 12 appearances for a Black Cats side that are 19th in the Premier League.<br>He was injured during his side's 3-0 win over Hull City on Saturday.<br>"We won't see him again this season and all we can hope is getting him right for the start of next season," said Sunderland boss David Moyes.<br>"I think he'd just started to find his way in the Premier League - even though he had experience at Manchester United - and the games he was having were bringing him on and giving him confidence."<br>The injury means McNair is unlikely to feature in Northern Ireland's World Cup 2018 qualifier against Norway in March.</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

sciq_pairs

• Dataset: sciq_pairs at 2c94ad3
• Size: 4,987 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 7 tokens</li><li>mean: 16.93 tokens</li><li>max: 76 tokens</li></ul>	<ul><li>min: 2 tokens</li><li>mean: 86.84 tokens</li><li>max: 512 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>The procedure known as angioplasty is used when what part of the circulatory system is clogged?</code>	<code>When a blood vessel gets clogged, there is no medical equivalent of "Drano" that will clear it out. There is, however, a procedure known as angioplasty. A thin tube with a balloon is threaded through the blood vessels. Once in place, the balloon is inflated to compress the clog against the artery wall.</code>
  <code>Mollusks such as squid and octopi, which must hunt to survive, possess what complex organs containing millions of neurons?</code>	<code>Figure 35.2 Nervous systems vary in structure and complexity. In (a) cnidarians, nerve cells form a decentralized nerve net. In (b) echinoderms, nerve cells are bundled into fibers called nerves. In animals exhibiting bilateral symmetry such as (c) planarians, neurons cluster into an anterior brain that processes information. In addition to a brain, (d) arthropods have clusters of nerve cell bodies, called peripheral ganglia, located along the ventral nerve cord. Mollusks such as squid and (e) octopi, which must hunt to survive, have complex brains containing millions of neurons. In (f) vertebrates, the brain and spinal cord comprise the central nervous system, while neurons extending into the rest of the body comprise the peripheral nervous system. (credit e: modification of work by Michael Vecchione, Clyde F. Roper, and Michael J. Sweeney, NOAA; credit f: modification of work by NIH).</code>
  <code>Combining nonpolar olive oil and polar vinegar yields what type of mixture?</code>	<code>Another familiar example is the mixing of vinegar and olive oil. Olive oil is a nonpolar substance, while vinegar (which is mostly water and acetic acid) is polar. The result is a heterogeneous mixture that exhibits a bilayer.</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

qasc_pairs

• Dataset: qasc_pairs at a34ba20
• Size: 4,987 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 4 tokens</li><li>mean: 11.33 tokens</li><li>max: 22 tokens</li></ul>	<ul><li>min: 13 tokens</li><li>mean: 33.52 tokens</li><li>max: 67 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>What happens before the rigid structure that surrounds the plant cell dilates?</code>	<code>Dilation occurs when cell walls relax.. Cell wall is the rigid structure that surrounds the plant cell. <br> Dilation occurs when the rigid structure that surrounds the plant cell relaxes.</code>
  <code>Reusing plastic bags has a positive impact on what?</code>	<code>recycling has a positive impact on the environment. Plastic bags are recyclable and they are reusable. <br> Reusing plastic bags has a positive impact on the environment</code>
  <code>What protects the body from harmful substances?</code>	<code>skin is used for protecting the body from harmful substances. Skin is a protective organ. <br> organs protect the body from harmful substances</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

openbookqa_pairs

• Dataset: openbookqa_pairs
• Size: 3,007 training samples
• Columns: <code>question</code> and <code>fact</code>
• Approximate statistics based on the first 1000 samples:

  	question	fact
  type	string	string
  details	<ul><li>min: 3 tokens</li><li>mean: 13.75 tokens</li><li>max: 78 tokens</li></ul>	<ul><li>min: 4 tokens</li><li>mean: 11.34 tokens</li><li>max: 31 tokens</li></ul>

• Samples:

  question	fact
  <code>Heat exposure at higher temperatures without ability to regulate internal body temperatures will result in the expiration of which of these?</code>	<code>if an organism becomes too hot then that organism may die</code>
  <code>Which of the following would be part of the water cycle?</code>	<code>evaporation is a stage in the water cycle process</code>
  <code>polar bears are white due to an inherited</code>	<code>the color of fur is an inherited characteristic</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

msmarco_pairs

• Dataset: msmarco_pairs at 28ff31e
• Size: 4,987 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 4 tokens</li><li>mean: 8.68 tokens</li><li>max: 21 tokens</li></ul>	<ul><li>min: 16 tokens</li><li>mean: 75.36 tokens</li><li>max: 238 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>how many days is it in a month</code>	<code>Each month has either 28, 30, or 31 days during a common year, which has 365 days. During leap years, which occur nearly every 4 years, we add an extra (intercalary) day, Leap Day, on 29 February, making leap years 366 days long.</code>
  <code>who were the peloponnesian wars between? who won them?</code>	<code>The Greek city-state of Sparta won the war against Athens. The war, known as the Peloponnesian War, raged for 27 years between the Athenian realm and the Peloponnesian coalition commanded by the Spartans. The Peloponnesian War began in 431 B.C.C. and ended in 404 B.C.E. when Athens conceded defeat to Sparta.</code>
  <code>average nurse practitioner salary ny</code>	<code>Nurse Practitioner New York, NY Salary. Nurse Practitioner New York, NY average salary is $91,897, median salary is $79,060 with a salary range from $20,530 to $1,926,393. Nurse Practitioner New York, NY salaries are collected from government agencies and companies. Each salary is associated with a real job position.</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

nq_pairs

• Dataset: nq_pairs at f9e894e
• Size: 4,987 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 10 tokens</li><li>mean: 11.79 tokens</li><li>max: 23 tokens</li></ul>	<ul><li>min: 14 tokens</li><li>mean: 131.1 tokens</li><li>max: 512 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>parks and rec episode where ron gets a hernia</code>	<code>The Stakeout (Parks and Recreation) Back at city hall, Ron (Nick Offerman) remains immobile in his chair all day due to a hernia which causes excruciating pain if he moves. Ron remains immobile in his seat well into the night, until the janitors turn the lights off on him. The intern, April (Aubrey Plaza), returns to check on him, and wheels Ron out to the car on his office chair to bring him to the hospital.</code>
  <code>when did jane beale come back to eastenders</code>	<code>Jane Beale Jane Beale (also Collins and Clarke) is a fictional character from the BBC soap opera, EastEnders, played by Laurie Brett. She made her first appearance on 21 June 2004. Brett took maternity leave in 2011[1] and departed on 19 May.[2] She returned on 8 November[3] and departed again on 27 January 2012.[4][5] Jane made temporary a return to the show on 6 January 2014 until 20 May and permanently from 24 November 2014.[6][7] Her major storylines have included her relationship and later marriages to Ian Beale (Adam Woodyatt); the first ended due to his affair with Glenda Mitchell (Glynis Barber), an affair with Grant Mitchell (Ross Kemp), accidentally shot by Ian's stepson Steven Beale (Aaron Sidwell) which results her desire to have a child of her own following a hysterectomy, a relationship with Masood Ahmed (Nitin Ganatra), her popular friendship with Tanya Branning (Jo Joyner), covering up her adopted son Bobby's (Eliot Carrington) role in the murder of her stepdaughter Lucy Beale (Hetti Bywater), being paralysed after brutally attacked with a hockey stick by Bobby, caught up in the fire at Beale's restaurant, which was started by Steven before she was then left in the blaze by Max Branning (Jake Wood) and being forced by Max to leave Walford which led to a second departure on 23 October 2017.[8] She made a voiceover appearance on 8 December.</code>
  <code>who sings what lovers do with maroon 5</code>	<code>What Lovers Do "What Lovers Do" is a song by American pop rock band Maroon 5 featuring American R&B singer SZA. It was released on August 30, 2017, as the lead single from the band's sixth studio album Red Pill Blues (2017).[4] The song contains an interpolation of the 2016 song "Sexual" by Neiked featuring Dyo, therefore Victor Rådström, Dyo and Elina Stridh are credited as songwriters.</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

trivia_pairs

• Dataset: trivia_pairs at a7c36e3
• Size: 4,987 training samples
• Columns: <code>query</code> and <code>answer</code>
• Approximate statistics based on the first 1000 samples:

  	query	answer
  type	string	string
  details	<ul><li>min: 8 tokens</li><li>mean: 17.26 tokens</li><li>max: 64 tokens</li></ul>	<ul><li>min: 24 tokens</li><li>mean: 205.76 tokens</li><li>max: 402 tokens</li></ul>

• Samples:

  query	answer
  <code>What word is the most common synonym of the word “corsair”?</code>	<code>Corsair Synonyms, Corsair Antonyms
  <code>"What scale initially related wind conditions to their effects on the sails of a man of war, from ""sufficient for steerage"" to ""which canvas sails couldn't withstand."""</code>	<code>Oil droplets transport due to irregular waves: Development of large-scale spreading coefficients "The sea state can be defined in a variety of ways. For example, one may use the Beaufort scale (Singleton, 2008) that describes the sea state based on qualitative terms. It is an empirical yet expedient scale. " [Show abstract] [Hide abstract] ABSTRACT: The movement of oil droplets due to waves and buoyancy was investigated by assuming an irregular sea state following a JONSWAP spectrum and four buoyancy values. A technique known as Wheeler stretching was used to model the movement of particles under the moving water surface. In each simulation, 500 particles were released and were tracked for a real time of 4.0 h. A Monte Carlo approach was used to obtain ensemble properties. It was found that small eddy diffusivities that decrease rapidly with depth generated the largest horizontal spreading of the plume. It was also found that large eddy diffusivities that decrease slowly with depth generated the smallest horizontal spreading coefficient of the plume. The increase in buoyancy resulted in a decrease in the horizontal spreading coefficient, which suggests that two-dimensional (horizontal) models that predict the transport of surface oil could be overestimating the spreading of oil. Full-text · Article · Jan 2016</code>
  <code>What is the name given to the collection of Welsh legends?</code>	<code>1000+ images about Welsh - Legends & Folklore on Pinterest

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

gooaq_pairs

• Dataset: gooaq_pairs at b089f72
• Size: 4,987 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 8 tokens</li><li>mean: 11.52 tokens</li><li>max: 19 tokens</li></ul>	<ul><li>min: 13 tokens</li><li>mean: 56.25 tokens</li><li>max: 143 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>how long does it take to get used to getting up earlier?</code>	<code>The most efficient method for changing the time you wake up is to do it gradually — 10–15 minutes earlier for 1–3 days until you feel used to it, and then lower it down. If you get up at 8 a.m. generally, don't suddenly change it to 6 a.m. Try 7:45 a.m. first.</code>
  <code>what are the differences among standards goals and objectives?</code>	<code>Standard: Written expectations of what students are expected to know at a specific stage of education. These are based off of learning objectives. Goal: Unique to an individual student. May stem from a standard, but also relates to a student's academic habits and organization.</code>
  <code>how do u work out the surface area of a triangular prism?</code>	<code>A triangular prism has three rectangular sides and two triangular faces. To find the area of the rectangular sides, use the formula A = lw, where A = area, l = length, and h = height. To find the area of the triangular faces, use the formula A = 1/2bh, where A = area, b = base, and h = height.</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

paws-pos

• Dataset: paws-pos at 161ece9
• Size: 4,987 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 8 tokens</li><li>mean: 25.62 tokens</li><li>max: 52 tokens</li></ul>	<ul><li>min: 8 tokens</li><li>mean: 25.61 tokens</li><li>max: 52 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>Primidone also causes exfoliative dermatitis , Johnson -- Stevens syndrome , and toxic epidermal necrolysis .</code>	<code>Primidone also causes exfoliative dermatitis , Johnson - Stevens -- Syndrome and toxic epidermal necrolysis .</code>
  <code>The agency was founded in 1976 in Chicago , and it entered the New York market in 1998 and Milwaukee in 2009 .</code>	<code>The agency was founded in Chicago in 1976 and entered New York in 1998 and Milwaukee in 2009 .</code>
  <code>After his death , the widow of Kellow Mary Hope Kellow with her daughter Mary moved to Sydney .</code>	<code>After his death , Kellow 's widow Mary Hope Kellow moved to Sydney with her daughter Mary .</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

global_dataset

• Dataset: global_dataset
• Size: 36,619 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 4 tokens</li><li>mean: 28.68 tokens</li><li>max: 368 tokens</li></ul>	<ul><li>min: 2 tokens</li><li>mean: 55.88 tokens</li><li>max: 512 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>The Turks , Tibetans , Muslim Arabs , and Tang competed for control of Central Asia until the tang 's collapse in the 10th century .</code>	<code>The Turks , Tang , Muslim Arabs and the Tibetans competed for control over Central Asia until the collapse of the Tang in the 10th century .</code>
  <code>What do animals use to reproduce?</code>	<code>an animal needs to attract a mate to reproduce. Animals mate because of smells. <br> animals attract with smells</code>
  <code>Some touch receptors sense a difference in pain or what?</code>	<code>Some touch receptors sense differences in temperature or pain.. Heat and temperature are the same. <br> Some touch receptors sense differences in heat or pain.</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

Evaluation Datasets

vitaminc-pairs

• Dataset: vitaminc-pairs at be6febb
• Size: 128 evaluation samples
• Columns: <code>claim</code> and <code>evidence</code>
• Approximate statistics based on the first 1000 samples:

  	claim	evidence
  type	string	string
  details	<ul><li>min: 9 tokens</li><li>mean: 19.71 tokens</li><li>max: 38 tokens</li></ul>	<ul><li>min: 9 tokens</li><li>mean: 32.5 tokens</li><li>max: 78 tokens</li></ul>

• Samples:

  claim	evidence
  <code>Dragon Con had over 5000 guests .</code>	<code>Among the more than 6000 guests and musical performers at the 2009 convention were such notables as Patrick Stewart , William Shatner , Leonard Nimoy , Terry Gilliam , Bruce Boxleitner , James Marsters , and Mary McDonnell .</code>
  <code>COVID-19 has reached more than 185 countries .</code>	<code>As of , more than cases of COVID-19 have been reported in more than 190 countries and 200 territories , resulting in more than deaths .</code>
  <code>In March , Italy had 3.6x times more cases of coronavirus than China .</code>	<code>As of 12 March , among nations with at least one million citizens , Italy has the world 's highest per capita rate of positive coronavirus cases at 206.1 cases per million people ( 3.6x times the rate of China ) and is the country with the second-highest number of positive cases as well as of deaths in the world , after China .</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

negation-triplets

• Dataset: negation-triplets
• Size: 128 evaluation samples
• Columns: <code>anchor</code>, <code>entailment</code>, and <code>negative</code>
• Approximate statistics based on the first 1000 samples:

  	anchor	entailment	negative
  type	string	string	string
  details	<ul><li>min: 8 tokens</li><li>mean: 14.54 tokens</li><li>max: 46 tokens</li></ul>	<ul><li>min: 6 tokens</li><li>mean: 12.12 tokens</li><li>max: 21 tokens</li></ul>	<ul><li>min: 6 tokens</li><li>mean: 12.43 tokens</li><li>max: 22 tokens</li></ul>

• Samples:

  anchor	entailment	negative
  <code>A Marine that is looking at his cell phone.</code>	<code>A marine in uniform using a smart phone.</code>	<code>Not a marine in uniform using a smart phone.</code>
  <code>A snowboarder on a wide plain of snow</code>	<code>A snow field with a snowboarder on it</code>	<code>An empty field with no snowboarder on it</code>
  <code>Three men, one holding pipes, another holding a large object above his head, and one resting against the pipe bed on the truck, are looking at the camera.</code>	<code>three men look at the camera</code>	<code>three men ignore the camera</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

scitail-pairs-pos

• Dataset: scitail-pairs-pos at 0cc4353
• Size: 128 evaluation samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 9 tokens</li><li>mean: 20.13 tokens</li><li>max: 53 tokens</li></ul>	<ul><li>min: 8 tokens</li><li>mean: 15.48 tokens</li><li>max: 23 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>humans normally have 23 pairs of chromosomes.</code>	<code>Humans typically have 23 pairs pairs of chromosomes.</code>
  <code>A solution is a homogenous mixture of two or more substances that exist in a single phase.</code>	<code>Solution is the term for a homogeneous mixture of two or more substances.</code>
  <code>Upwelling The physical process in near-shore ocean systems of rising of nutrients and colder bottom waters to the surface because of constant wind patterns along the shoreline.</code>	<code>Upwelling is the term for when deep ocean water rises to the surface.</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

scitail-pairs-qa

• Dataset: scitail-pairs-qa at 0cc4353
• Size: 128 evaluation samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 7 tokens</li><li>mean: 16.6 tokens</li><li>max: 33 tokens</li></ul>	<ul><li>min: 8 tokens</li><li>mean: 15.67 tokens</li><li>max: 33 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>Magma comes toward earth's crust through mantle plumes.</code>	<code>What substance comes toward earth's crust through mantle plumes?</code>
  <code>The understory of the rainforest commonly has ferns and other ground plants.</code>	<code>What part of the rainforest commonly has ferns and other ground plants?</code>
  <code>Because trees add water vapor to air, cutting down forests leads to longer periods of drought.</code>	<code>Because trees add water vapor to air, cutting down forests leads to longer periods of what?</code>

• Loss: <code>CachedGISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

xsum-pairs

• Dataset: xsum-pairs
• Size: 128 evaluation samples
• Columns: <code>summary</code> and <code>document</code>
• Approximate statistics based on the first 1000 samples:

  	summary	document
  type	string	string
  details	<ul><li>min: 13 tokens</li><li>mean: 25.42 tokens</li><li>max: 46 tokens</li></ul>	<ul><li>min: 67 tokens</li><li>mean: 213.0 tokens</li><li>max: 354 tokens</li></ul>

• Samples:

  summary	document
  <code>An exam paper for Scottish secondary pupils is being re-issued after a mistake was spotted.</code>	<code>The National 5 Modern Studies exam will take place on Friday afternoon.<br>The Scottish Qualification Authority (SQA) said the paper has been reprinted because of a typographical error.<br>The exams body said there was no suggestion of a security breach. A spokesman added that it had procedures in place to deal with situations like this.<br>The printing mistake in the original paper was in a diagram, not a question.<br>The SQA said it acted quickly after the error was spotted.<br>The reprinted paper - with the correct text in the diagram but otherwise identical - will be issued to exam centres across Scotland.<br>Last year the Higher English paper was replaced amid fears of a security breach.<br>The SQA also faced criticism over mistakes in the National 5 computing paper.<br>Earlier this month, the SQA announced that teachers would not have access to exam papers until the day after each test takes place.<br>It said the change was in order to improve security and confidentiality. The largest teachers' union, the EIS, has condemned the move.</code>
  <code>US regulators have told seven carmakers the recall of airbags made by Japanese firm Takata is likely to expand.</code>	<code>The National Highway Traffic Safety Administration (NHTSA) has written to firms including Mercedes-Benz, Jaguar-Land Rover and Tesla to ask which of their models use the Takata parts.<br>About 23.4 million Takata airbag inflators have been recalled in the US.<br>The airbags have been linked to eight deaths and more than 100 injuries around the world.<br>It was found they can inflate with excessive force, spraying metal shrapnel at the drivers.<br>The driver and passenger airbags were in more than 19 million cars sold by 11 different companies such as Honda in the US.<br>In the letters sent last week, the NHTSA said the recall "will likely grow to include vehicles that are outside the scope of the current recalls".<br>The agency will attend a public meeting in Washington on 22 October to discuss the Takata investigation and whether it will take over management of the recalls to speed up the repairs.<br>Carmakers are struggling to get parts with only 4.4 million airbag inflators replaced since the start of this month.<br>The other automakers that received the letters include Suzuki, Volvo Trucks, Volkswagen and Spartan Motors.<br>So far Mercedes, Jaguar-Land Rover and Tesla have all said the air bags they used from Takata are not part of current recalls, according to the Associated Press.</code>
  <code>A salt lake in Melbourne has turned pink due to a combination of sunlight, warm temperatures and low rainfall.</code>	<code>Wildlife officers said algae growing in the salt crust at the bottom of Westgate Park's lake produce a red pigment.<br>"Enjoy the views, but we recommend you don't come into contact with the water," Parks Victoria said.<br>The phenomenon also occurs in Spain's Salina de Torrevieja, Canada's Dusty Rose Lake and Senegal's Lake Retba.<br>In Australia, the natural occurring sight can be seen in Victoria's Murray-Sunset National Park and Western Australia's Lake Hillier.<br>You might also be interested in:<br>Dr Mark Norman, Parks Victoria chief conservation scientist, said the colouration was caused by a harmless, single-cell alga known as Dunalliela.<br>"It's completely natural," he said. "We often get comments that it looks like an industrial accident of pink paint."<br>Dr Norman said that even though the water is not dangerous, he would not recommend taking a swim.<br>"It's so salty and muddy on the bottom that you would come out looking like a frosted rum ball, especially when you dried," he said.<br>Parks Victoria said the lake is expected to return to blue when the weather cooled and the rainfall increased.</code>

The model card is truncated. Read the rest on Hugging Face.