DeBERTa-small-ST-v1-test-step2

SentenceTransformer based on bobox/DeBERTa-small-ST-v1-test

This is a sentence-transformers model finetuned from bobox/DeBERTa-small-ST-v1-test 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 and paws-pos datasets. 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
• Base model: bobox/DeBERTa-small-ST-v1-test <!-- at revision 55585e7a71c7a0177a0a8a60c64ab08c0a5f84e3 -->
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 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
• 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': 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})
)

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/DeBERTa-small-ST-v1-test-step2")
# Run inference
sentences = [
    'where did the ice storm of 1998 happen',
    'January 1998 North American ice storm The North American Ice Storm of 1998 (also known as Great Ice Storm of 1998) was a massive combination of five smaller successive ice storms in January 1998 that struck a relatively narrow swath of land from eastern Ontario to southern Quebec, New Brunswick and Nova Scotia in Canada, and bordering areas from northern New York to central Maine in the United States. It caused massive damage to trees and electrical infrastructure all over the area, leading to widespread long-term power outages. Millions were left in the dark for periods varying from days to several weeks, and in some instances, months. It led to 35 fatalities, a shutdown of activities in large cities like Montreal and Ottawa, and an unprecedented effort in reconstruction of the power grid. The ice storm led to the largest deployment of Canadian military personnel since the Korean War, with over 16,000 Canadian Forces personnel deployed, 12,000 in Quebec and 4,000 in Ontario at the height of the crisis.[1][2]:16',
    'Nimbostratus cloud Nimbostratus is a stratiform genus formerly classified as "Family C" low-level, but now considered by the World Meteorological Organization (WMO) to be a middle- or multi-level stratus type.[1]. Although it is usually a low-based cloud, it actually forms most commonly in the middle level of the troposphere and then spreads vertically into the low and high levels. This change in classification would once have made it a "Family D" cloud, but this style of nomenclature was discontinued by the WMO in 1956. Nimbostratus usually produces precipitation over a wide area. Nimbo- is from the Latin word nimbus, which denotes precipitation. It has a diffuse cloud base generally found anywhere from near surface in the low levels to about 3,000\xa0m (9,800\xa0ft) in the middle level of the troposphere. Although usually dark at its base, it often appears illuminated from within to a surface observer.[2] Nimbostratus usually has a thickness of about 2000 m. Though found worldwide, nimbostratus occurs more commonly in the middle latitudes.[3] It is coded CM2 on the SYNOP report.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# 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: VitaminC
• 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: 39,000 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: 5 tokens</li><li>mean: 22.25 tokens</li><li>max: 372 tokens</li></ul>	<ul><li>min: 3 tokens</li><li>mean: 13.64 tokens</li><li>max: 49 tokens</li></ul>	<ul><li>min: 3 tokens</li><li>mean: 14.02 tokens</li><li>max: 49 tokens</li></ul>

• Samples:

  anchor	entailment	negative
  <code>A young man in white is midair on his bicycle performing a trick.</code>	<code>A young man in white is midair</code>	<code>A young man in white is not midair</code>
  <code>A bicycle has a red umbrella attached to it.</code>	<code>A parked bicycle with a red umbrella attached to it.</code>	<code>A parked bicycle without a red umbrella attached to it.</code>
  <code>Tanzania started rationing electricity after a technical problem shut down machinery at the Songas gas-fired power plant in Dar es Salaam, the state-run power utility said.</code>	<code>Tanzania rations electricity after technical problem at plant</code>	<code>Tanzania boosts electricity after upgrade at plant</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': 0.05}
  

vitaminc-pairs

• Dataset: vitaminc-pairs at be6febb
• Size: 36,000 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: 7 tokens</li><li>mean: 17.47 tokens</li><li>max: 55 tokens</li></ul>	<ul><li>min: 8 tokens</li><li>mean: 36.01 tokens</li><li>max: 164 tokens</li></ul>

• Samples:

  claim	evidence
  <code>Candice Crawford participated in 2 beauty pageants .</code>	<code>He has a younger sister , former Miss Missouri USA winner and [ [ Miss USA ] contestant Candice Crawford .</code>
  <code>Rio Ferdinand plays for Queens Park Rangers on a free transfer .</code>	<code>A centre-back , he is currently plays for Queens Park Rangers on a Free Transfer after leaving Manchester United following twelve years at the club .</code>
  <code>Matt Damon is credited in Deadpool 2 as Dickie Grrenleaf , in reference to Jude Law 's character from The Talented Mr. Ripley .</code>	<code>This was inspired by a real manifesto written by Reese , which the writers wanted to be discussed in the film by a certain calibre '' of actor : the characters are portrayed by Alan Tudyk and a disguised Matt Damon , with the latter credited as Dickie Greenleaf '' ( a reference to Jude Law 's character from the 1999 film The Talented Mr. Ripley in which Damon stars ) .</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': 0.05}
  

scitail-pairs-qa

• Dataset: scitail-pairs-qa at 0cc4353
• Size: 14,237 training samples
• Columns: <code>sentence2</code> and <code>sentence1</code>
• Approximate statistics based on the first 1000 samples:

  	sentence2	sentence1
  type	string	string
  details	<ul><li>min: 7 tokens</li><li>mean: 15.9 tokens</li><li>max: 41 tokens</li></ul>	<ul><li>min: 6 tokens</li><li>mean: 15.07 tokens</li><li>max: 41 tokens</li></ul>

• Samples:

  sentence2	sentence1
  <code>How ice cores are important to the study of geologic history is best described as they contain evidence showing changes in the atmospheric composition over time.</code>	<code>Which best describes how ice cores are important to the study of geologic history?</code>
  <code>The lens of the eye is a(n) convex shape.</code>	<code>What shape is the lens of the eye?</code>
  <code>Most ferns produce the same type of spores and are therefore called homosporous.</code>	<code>Most ferns produce the same type of spores and are therefore called 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': 0.05}
  

scitail-pairs-pos

• Dataset: scitail-pairs-pos at 0cc4353
• Size: 8,600 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: 24.17 tokens</li><li>max: 69 tokens</li></ul>	<ul><li>min: 7 tokens</li><li>mean: 15.66 tokens</li><li>max: 37 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>The rate of decay is conveniently expressed in terms of an isotope's half-life, or the time it takes for one-half of a particular radioactive isotope in a sample to decay.</code>	<code>The term half-life decribes the amount of time required for half of the original material to decay in an isotope.</code>
  <code>This rock, called magma, furnishes the heat for the park's geysers and hot springs.</code>	<code>The water in some springs are hot because they're heated by hot magma.</code>
  <code>Carbon, with four valence electrons, forms covalent bonds to four neighboring carbon atoms arranged toward the corners of a tetrahedron, as shown in the figure below.</code>	<code>Four valence electrons can be found in a carbon atom.</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': 0.05}
  

xsum-pairs

• Dataset: xsum-pairs
• Size: 36,000 training samples
• Columns: <code>document</code> and <code>summary</code>
• Approximate statistics based on the first 1000 samples:

  	document	summary
  type	string	string
  details	<ul><li>min: 56 tokens</li><li>mean: 219.62 tokens</li><li>max: 371 tokens</li></ul>	<ul><li>min: 12 tokens</li><li>mean: 25.36 tokens</li><li>max: 56 tokens</li></ul>

• Samples:

  document	summary
  <code>Christopher Paul O'Kane, aged 42, from Woodland Avenue, was due to stand trial on Wednesday on 17 Provisional IRA related terrorist offences, including the attempted murder of a police officer more than 20 years ago.<br>However, O'Kane replied "guilty'' to five of the five terror charges when they were put to him in court.<br>These were: assisting an offender in relation to the murder of Constable Michael Ferguson in January 1994, and firing a mortar bomb at a police vehicle in Fanad Drive in October 1993. Planting a bomb at the railway line at Ebrington Barracks in December 1993; placing a bomb at the home of a senior police officer at Prehen in 1994, and planting a bomb at Forge George army base in Derry.<br>Prosecution lawyer Robin Steer asked that the remaining 12 charges "be left on the books" and not to be proceeded without the leave of the Court or the Court of Appeal.<br>O'Kane will be sentenced on December 16.</code>	<code>A Londonderry man has pleaded guilty at Belfast Crown Court to five terrorist offences relating to bomb attacks against security force members.</code>
  <code>Chase, 31, has made five appearances for the Vikings since the initial temporary move, having been told he was not part of Castleford's 2018 plans.<br>In addition, Widnes have signed hooker Danny Walker to a new four-year contract after his first-grade breakthrough this season.<br>Head coach Denis Betts said: "The deals represent a real statement of intent."<br>New Zealand-born Chase was the 2011 Man of Steel, has played for the Tigers, Salford, Leigh and now Widnes, as well as representing England at senior level.<br>In contrast, Walker, 18, is at the opposite end of his career, with just six professional appearances to his name.<br>"Rangi has made a big impact since his arrival, contributing well both on and off the field," Betts added.<br>"He has exceptional abilities and, as an international standard half-back, adds real quality to the team.<br>"Danny is one of the most promising young players in the Super League. His commitment to a four-year contract shows his belief in what we are building at the Vikings."</code>	<code>Widnes have signed half-back Rangi Chase from Castleford on a permanent deal after a successful loan spell.</code>
  <code>The characters include Nebula, a blue skinned alien played by Scottish actress Karen Gillan.<br>There was a backlash on social media to merchandise released for the first Guardians film in 2014.<br>Gamora, who is played by Zoe Saldana, did not appear on a t-shirt while her fellow, male leads did.<br>The row saw the hashtag #WheresGamora appear on Twitter.<br>Fans of the film also complained that other merchandise of Gamora, and Gillan's Nebula, who also appeared in the first movie, was harder to find than that featuring male characters.<br>In a post on Facebook on Sunday, the films' director James Gunn confirmed that there was a commitment to raising the profile of the female characters.<br>He wrote: "Guardians of the Galaxy Vol. 2 will have Mantis, Nebula, Gamora and Elizabeth Debicki's character as part of the fray, and we're committed to making sure they're included in more toys and merch than the last go round."<br>Marvel and Disney are involved in making the movies and commissioning the merchandising.<br>Guardians, which topped the US box office in 2014 with ticket sales of $332.8m (£203.9m), also stars Chris Pratt and Dave Bautista and the voices of Vin Diesel and Bradley Cooper.<br>Earlier this year, Gunn revealed that Gillan, from Inverness, has a bigger role in the new movie, which is to be released next year.</code>	<code>Female stars of the new Guardians of the Galaxy movie will feature more prominently in its merchandising, the film's makers have said.</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': 0.05}
  

sciq_pairs

• Dataset: sciq_pairs at 2c94ad3
• Size: 11,095 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: 6 tokens</li><li>mean: 16.94 tokens</li><li>max: 68 tokens</li></ul>	<ul><li>min: 2 tokens</li><li>mean: 80.74 tokens</li><li>max: 512 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>What branch of science is defined as the study of matter?</code>	<code>1.8 End-of-Chapter Material Chapter Summary To ensure that you understand the material in this chapter, you should review the meanings of the bold terms in the following summary and ask yourself how they relate to the topics in the chapter. Chemistry is the study of matter, which is anything that has mass and takes up space. Chemistry is one branch of science, which is the study of the natural universe. Like all branches of science, chemistry relies on the scientific method, which is a process of Saylor URL: http://www. saylor. org/books.</code>
  <code>The net effect of aldosterone is to conserve and increase water levels in the plasma by reducing the excretion of what element, and thus water?</code>	<code>Aldosterone Recall that aldosterone increases the excretion of potassium and the reabsorption of sodium in the distal tubule. Aldosterone is released if blood levels of potassium increase, if blood levels of sodium severely decrease, or if blood pressure decreases. Its net effect is to conserve and increase water levels in the plasma by reducing the excretion of sodium, and thus water, from the kidneys. In a negative feedback loop, increased osmolality of the ECF (which follows aldosterone-stimulated sodium absorption) inhibits the release of the hormone (Figure 26.13).</code>
  <code>What do all chemical sections need to get started?</code>	<code>The bonds between the atoms need to be rearranged. That is the definition of a chemical reaction. And all chemical sections need energy to get started.</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': 0.05}
  

qasc_pairs

• Dataset: qasc_pairs at a34ba20
• Size: 7,727 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: 5 tokens</li><li>mean: 11.21 tokens</li><li>max: 26 tokens</li></ul>	<ul><li>min: 15 tokens</li><li>mean: 34.56 tokens</li><li>max: 67 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>vertebrates have a complete digestive system and a what?</code>	<code>Chordates have a complete digestive system and a closed circulatory system.. Vertebrates are members of of a larger group, the chordates .. Vertebrates have a complete digestive system and a closed circulatory system</code>
  <code>What is plasma?</code>	<code>plasma is formed by electrons separating from atoms in stars. Stars shine because they are hot .. plasma is hot</code>
  <code>Tungsten filaments sealed in a glass bulb are used for what in the dark?</code>	<code>a light bulb is used for seeing in the dark. Incandescent light bulbs use a tungsten filament sealed inside a glass bulb.. Tungsten filament sealed in a glass bulb is used for seeing in the dark.</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': 0.05}
  

openbookqa_pairs

• Dataset: openbookqa_pairs
• Size: 4,522 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: 13.8 tokens</li><li>max: 78 tokens</li></ul>	<ul><li>min: 4 tokens</li><li>mean: 11.5 tokens</li><li>max: 30 tokens</li></ul>

• Samples:

  question	fact
  <code>What is animal competition?</code>	<code>if two animals eat the same prey then those animals compete for that pey</code>
  <code>If you wanted to make a metal bed frame, where would you start?</code>	<code>alloys are made of two or more metals</code>
  <code>Places lacking warmth have few what</code>	<code>cold environments contain few organisms</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': 0.05}
  

msmarco_pairs

• Dataset: msmarco_pairs at 28ff31e
• Size: 33,000 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: 8.45 tokens</li><li>max: 34 tokens</li></ul>	<ul><li>min: 16 tokens</li><li>mean: 76.9 tokens</li><li>max: 215 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>what is the salary of hvac service mana</code>	<code>According to the Bureau of Labor Statistics, the median salary for HVAC technician in 2015 is $45,110 per year and approximately $21,69 per hour. Salary variations according to various factors Depending on technician’s ...</code>
  <code>caffeine effects on nervous system</code>	<code>Caffeine is a central nervous system stimulant that reduces fatigue and drowsiness. At normal doses, caffeine has variable effects on learning and memory, but it generally improves reaction time, wakefulness, concentration, and motor coordination.</code>
  <code>cost to flush transmission</code>	<code>1 Flushing a transmission with a pressurized machine can cost $125-$300 or more, and may include pushing a special cleaning compound through the system. 2 The power flush process typically replaces all of the transmission fluid, and can require 12-22 or more quarts.</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': 0.05}
  

nq_pairs

• Dataset: nq_pairs at f9e894e
• Size: 33,000 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: 11.82 tokens</li><li>max: 25 tokens</li></ul>	<ul><li>min: 14 tokens</li><li>mean: 135.44 tokens</li><li>max: 512 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>is watcher in the woods a disney movie</code>	<code>The Watcher in the Woods Filmed at Pinewood Studios and the surrounding areas in Buckinghamshire, England, The Watcher in the Woods was one of several live-action films produced by Walt Disney Productions in the 1980s, when the studio was targeting young adult audiences. The film suffered from various production problems and was pulled from theatres after its initial release in 1980. It was re-released in 1981 after being re-edited and a revised ending added.</code>
  <code>how much money did the ice bucket challenge raise for als</code>	<code>Ice Bucket Challenge Within weeks of the challenge going viral, The New York Times reported that the ALS Association had received $41.8 million in donations from more than 739,000 new donors from July 29 until August 21, more than double the $19.4 million the association received during the year that ended January 31, 2013.[89] On August 29, the ALS Association announced that their total donations since July 29 had exceeded $100 million.[90] The ALS Association is just one of several ALS-related charities that have benefited from the challenge:</code>
  <code>last episode of one foot in the grave</code>	<code>Things Aren't Simple Any More "Things Aren't Simple Any More" is the final episode of the British television sitcom One Foot in the Grave. It was written by David Renwick and stars Richard Wilson as Victor Meldrew, Annette Crosbie as his wife Margaret, and features guest appearances by Hannah Gordon and Paul Merton. The episode depicts the death of the series' protagonist, Victor Meldrew, in a hit-and-run road accident, and his wife's efforts to deal with the driver who killed him. Renwick had been struggling to conceive and write new stories for the series and decided to kill off the character. The episode was filmed at Shawford, Hampshire, and at BBC Television Centre in London.[3]</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': 0.05}
  

trivia_pairs

• Dataset: trivia_pairs at a7c36e3
• Size: 30,000 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: 18.69 tokens</li><li>max: 97 tokens</li></ul>	<ul><li>min: 36 tokens</li><li>mean: 458.53 tokens</li><li>max: 512 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>Which Channel Four game show was remade in the USA under the title Junkyard Wars?</code>	<code>Junkyard Wars Needs A Few Good Contestants - Slashdot Slashdot Follow Slashdot stories on Twitter   Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. × 1380727 story on Saturday January 27, 2001 @09:07AM from the gentlemen-start-your-torches dept. Andy B writes: "At long last, we have got an address for aspiring contestants to send their Junkyard Wars applications to (Slashdot slightly jumped the gun last autumn). Hurry to get you applications in, as the deadline for applications is Fri 16th Feb." ← You may like to read: → Yes, but its license states that it must act as a public service broadcaster: This [itc.org.uk] is taken from the ITC [itc.org.uk] website: The main points in the new licence are: redefinition of the remit in relation to all channels, not just ITV, and further commitment to innovation and experiment; a revised and strengthened statement on education; a commitment to provide at least three hours on average per week of multicultural programmes, and also to schedule at least some of these in peak time; a major commitment to the UK film industry, giving some preference to innovative and risky subjects and treatments; a new commitment to programmes for and about people with disabilities; increased requirement of 60 per cent of programmes specially commissioned for the Channel by 1999; a new commitment for production outside the London region, including a minimum requirement of 30 per cent by 2002; new maxima for repeats; a new commitment and new minimum requirement for spending on training; a new requirement for diversity in the peak-time schedule, including news, current affairs, educational, religious and multicultural programmes; revised commitments to subtitling and other provisions for those with hearing and sight impairments. I think it's not fully privatised either - doesn't the government still own a proportion of it and fund it a bit too? by MrP- ( 45616 ) writes: if you watched the last episode of junkyard wars (the rocket one), cathy said you can go to the site and submit an idea for them to build, i forget what site she said, probably either the junkyard wars site, or tlc.com, either way you should submit that, that would be cool to see... but only thing is theyd probably want to fight during the show at the end, unless they go tape the fight at battlebots, show the match during junkyard wars, then battle bots later, like they did with jay lenos chinkilla, they aired the match early on the tonight show then again on battle bots... but i dont know since battle bots is comedy centrals, maybe theyd have to do robot wars, but they dont air robot wars enough in america :( by Uberminky ( 122220 ) writes: The new Junkyard Wars episodes stink. I mean yeah, it's still a cool show, but it's just not what it used to be. I miss the silly Brits, they cracked me up. But yunno... maybe it's just me, but I swear the type of things they do on the show have changed since they got the new host and stuff. I mean.. one of the last episodes the Scrapheap Challenge did was drag racing. And what's the first Junkyard Wars thing? Drag racing. They also did all-terrain vehicles, which Scrapheap Challenge did.. and.. just tons of them. They're all the same sorts of things. Maybe it's just my perception, and the fact that I miss the old show, but it really seems like they've dumbed it down to suit us redneck Americans or something. I dunno... I just wish they'd give me my Scrapheap Challenge... Looking forward to next week's episode.... by ScuzzMonkey ( 208981 ) writes: I haven't seen all that many episodes, but of those I have seen, this seems to be the theme. The team that comes up with the more brilliant, elegant design has it crap out on them almost immediately, because, after all, it's made out of junk. The crowd that bangs together some brute-force job powers through and wins it. Apparently, it doesn't pay to get too creative. Makes it more fun to watch, though. by Uberminky ( 122220 ) writes: that would be so lame. On the one hand, we have people turning scrap met</code>
  <code>Which 2010 film stars Mila Kunis as Lily and Natalie Portman as Nina Sayers?</code>	<code>Black Swan (2010) - IMDb IMDb There was an error trying to load your rating for this title. Some parts of this page won't work property. Please reload or try later. X Beta I'm Watching This! Keep track of everything you watch; tell your friends. Error From $2.99 (SD) on Amazon Video ON DISC A committed dancer wins the lead role in a production of Tchaikovsky's "Swan Lake" only to find herself struggling to maintain her sanity. Director: a list of 34 titles created 25 Dec 2012 a list of 27 images created 09 Mar 2013 a list of 43 titles created 10 Nov 2013 a list of 26 titles created 05 Nov 2014 a list of 48 titles created 14 Mar 2015 Search for " Black Swan " on Amazon.com Connect with IMDb Want to share IMDb's rating on your own site? Use the HTML below. You must be a registered user to use the IMDb rating plugin. Won 1 Oscar. Another 90 wins & 245 nominations. See more awards  » Videos Harvard student Mark Zuckerberg creates the social networking site that would become known as Facebook, but is later sued by two brothers who claimed he stole their idea, and the co-founder who was later squeezed out of the business. Director: David Fincher Two astronauts work together to survive after an accident which leaves them alone in space. Director: Alfonso Cuarón After a stint in a mental institution, former teacher Pat Solitano moves back in with his parents and tries to reconcile with his ex-wife. Things get more challenging when Pat meets Tiffany, a mysterious girl with problems of her own. Director: David O. Russell A Mumbai teen reflects on his upbringing in the slums when he is accused of cheating on the Indian Version of "Who Wants to be a Millionaire?" Directors: Danny Boyle, Loveleen Tandan Stars: Dev Patel, Freida Pinto, Saurabh Shukla A young man who survives a disaster at sea is hurtled into an epic journey of adventure and discovery. While cast away, he forms an unexpected connection with another survivor: a fearsome Bengal tiger. Director: Ang Lee The story of King George VI of the United Kingdom of Great Britain and Northern Ireland, his impromptu ascension to the throne and the speech therapist who helped the unsure monarch become worthy of it. Director: Tom Hooper In 1985 Dallas, electrician and hustler Ron Woodroof works around the system to help AIDS patients get the medication they need after he is diagnosed with the disease. Director: Jean-Marc Vallée A paraplegic marine dispatched to the moon Pandora on a unique mission becomes torn between following his orders and protecting the world he feels is his home. Director: James Cameron A lonely writer develops an unlikely relationship with an operating system designed to meet his every need. Director: Spike Jonze Tells the story of Benjamin Button, a man who starts aging backwards with bizarre consequences. Director: David Fincher A family determined to get their young daughter into the finals of a beauty pageant take a cross-country trip in their VW bus. Directors: Jonathan Dayton, Valerie Faris Stars: Steve Carell, Toni Collette, Greg Kinnear A seventeen-year-old aristocrat falls in love with a kind but poor artist aboard the luxurious, ill-fated R.M.S. Titanic. Director: James Cameron Edit Storyline Nina (Portman) is a ballerina in a New York City ballet company whose life, like all those in her profession, is completely consumed with dance. She lives with her obsessive former ballerina mother Erica (Hershey) who exerts a suffocating control over her. When artistic director Thomas Leroy (Cassel) decides to replace prima ballerina Beth MacIntyre (Ryder) for the opening production of their new season, Swan Lake, Nina is his first choice. But Nina has competition: a new dancer, Lily (Kunis), who impresses Leroy as well. Swan Lake requires a dancer who can play both the White Swan with innocence and grace, and the Black Swan, who represents guile and sensuality. Nina fits the White Swan role perfectly but Lily is the personification of the Black Swan. As the two young dancers expand their rivalry into a twisted friendship, Nina begins to</code>
  <code>In computing, what is the device which is plugged into a computer which serves as an adapter or to enable the use of certain software?</code>	<code>What Is A Dongle? - Business Insider print 3 Dongles plugged into Google's Chromebook Pixel. Kevin Smith/Business Insider This week two people in the tech industry lost their jobs because of jokes gone awry at a conference. A couple of male developers were joking around about "big" dongles. When a female developer evangelist heard their jokes she tweeted it out with their photo, complaining that they were being rude. One of those developers lost his job, and then eventually, so did the evangelist. In the developers' defense, the word "dongle" is funny. It's nearly impossible to say without giggling or making childish jokes. Some people in our office had heard the word before, but didn't know what a dongle is. Merriam-Webster  defines a dongle as a small device that plugs into a computer and serves as an adapter or a security measure to enable the use of certain software. Kevin Smith/Business Insider The term, dongle, was rumored to have originated from a 1992 advertisement for Rainbow Technologies. The ad claimed the word dongle was derived from the name "Don Gall." Though untrue, this has given rise to an urban myth, we learned from Wikipedia . According to the University of Pennsylvania's language log , the earliest citation of the word dongle began appearing in 1982: 1982   MicroComputer Printout  Jan. 19/2 The word ‘dongle’ has been appearing in many articles with reference to security systems for computer software [refers to alleged coinage in 1980]. But as the term 'dongle' became more widespread its meaning changed from strictly a scientific term to mainstream. UPenn clarifies: The current meaning for dongle seems to be something like "a self-contained device that plugs into a  port on a computer that is normally used for connections to a separate external device". Thus in addition to the original serial-port dongles, and the USB dongles that Suzanne (and Stephen Fry) wrote about, there are also  "firewire dongles" , and presumably there could be dongles for any other sort of port as well.  Simply put, dongles are computer peripherals that plug into your computer like a USB flash drive or a cord connecting a computer with a printer for example.  Dongles are also huge in the video game world because they allow consoles to have added features like increased audio quality. An Xbox 360 Audio Dongle Wikimedia Commons Before USB was the standard in attaching PC peripherals, there were tons of different dongles that came in various shapes and sizes.</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': 0.05}
  

gooaq_pairs

• Dataset: gooaq_pairs at b089f72
• Size: 30,000 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: 11.43 tokens</li><li>max: 21 tokens</li></ul>	<ul><li>min: 14 tokens</li><li>mean: 57.57 tokens</li><li>max: 155 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>what is the difference between back off and back down?</code>	<code>You back off by retreating from a confrontation of some kind, but here the other person is not being antagonistic. You back down by giving way when you realise you're losing an argument or fight, but that doesn't apply here either.</code>
  <code>how many days after lh surge should i have intercourse?</code>	<code>The three days immediately after a positive test represent the best time to have intercourse to increase the probability of getting pregnant. Ovulation generally occurs a day or two after the LH surge.</code>
  <code>what episode does jane and rafael do it?</code>	<code>Rodriguez's directorial debut was the tenth episode of the season, "Chapter Seventy-Four", which aired on February 9, 2018. The episode was specially selected by showrunner Jennie Snyder-Urman as it featured the characters of Jane (Gina Rodriguez) and Rafael (Justin Baldoni) having sex for the first time.</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': 0.05}
  

paws-pos

• Dataset: paws-pos at 161ece9
• Size: 21,829 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: 9 tokens</li><li>mean: 25.51 tokens</li><li>max: 59 tokens</li></ul>	<ul><li>min: 8 tokens</li><li>mean: 25.47 tokens</li><li>max: 57 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>Mouhoun is one of the 45 provinces of Boucle du Mouhoun Region and is in Burkina Faso . The capital of Mouhoun is Dédougou .</code>	<code>Mouhoun is one of 45 provinces in the Boucle du Mouhoun region and is located in Burkina Faso , the capital of Mouhoun is Dédougou .</code>
  <code>Besides Quintin , they had five children : Juan , Phillip , Willie , Patrick and Lucy .</code>	<code>They had five children besides Quintin : Lucy , Phillip , Juan , Patrick and Willie .</code>
  <code>According to the U.S. Census Bureau , the county is a total area that has land and ( 0.2 % ) of water .</code>	<code>According to the US Census Bureau , the county has a total area of which is land and ( 0.2 % ) of water .</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': 0.05}
  

Evaluation Datasets

vitaminc-pairs

• Dataset: vitaminc-pairs at be6febb
• Size: 108 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: 21.36 tokens</li><li>max: 41 tokens</li></ul>	<ul><li>min: 11 tokens</li><li>mean: 36.11 tokens</li><li>max: 79 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': 0.05}
  

negation-triplets

• Dataset: negation-triplets
• Size: 64 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: 10 tokens</li><li>mean: 13.83 tokens</li><li>max: 19 tokens</li></ul>	<ul><li>min: 9 tokens</li><li>mean: 13.23 tokens</li><li>max: 19 tokens</li></ul>	<ul><li>min: 10 tokens</li><li>mean: 13.61 tokens</li><li>max: 19 tokens</li></ul>

• Samples:

  anchor	entailment	negative
  <code>Boy rides skateboard and does trick over stairs.</code>	<code>Man in air on skateboard at night by light of street lamps.</code>	<code>Man on the ground on skateboard during the day by light of street lamps.</code>
  <code>A small child climbs atop a large motorcycle</code>	<code>A young boy riding a motorcycle next to a silver car.</code>	<code>An adult riding a motorcycle next to a silver car.</code>
  <code>A group of people skiing down a snow covered slope.</code>	<code>Six people in snow field with ski equipment.</code>	<code>Six people in a desert with no ski equipment.</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': 0.05}
  

scitail-pairs-pos

• Dataset: scitail-pairs-pos at 0cc4353
• Size: 54 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: 20.81 tokens</li><li>max: 45 tokens</li></ul>	<ul><li>min: 10 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': 0.05}
  

xsum-pairs

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

  	document	summary
  type	string	string
  details	<ul><li>min: 51 tokens</li><li>mean: 219.56 tokens</li><li>max: 341 tokens</li></ul>	<ul><li>min: 15 tokens</li><li>mean: 25.73 tokens</li><li>max: 47 tokens</li></ul>

• Samples:

  document	summary

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