README.md

metadata

base_model: microsoft/deberta-v3-small
datasets:
  - jinaai/negation-dataset-v2
  - tals/vitaminc
  - allenai/scitail
  - allenai/sciq
  - allenai/qasc
  - sentence-transformers/msmarco-msmarco-distilbert-base-v3
  - sentence-transformers/natural-questions
  - sentence-transformers/trivia-qa
  - sentence-transformers/gooaq
  - google-research-datasets/paws
language:
  - en
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
  - pearson_manhattan
  - spearman_manhattan
  - pearson_euclidean
  - spearman_euclidean
  - pearson_dot
  - spearman_dot
  - pearson_max
  - spearman_max
  - cosine_accuracy
  - dot_accuracy
  - manhattan_accuracy
  - euclidean_accuracy
  - max_accuracy
  - cosine_accuracy_threshold
  - cosine_f1
  - cosine_f1_threshold
  - cosine_precision
  - cosine_recall
  - cosine_ap
  - dot_accuracy_threshold
  - dot_f1
  - dot_f1_threshold
  - dot_precision
  - dot_recall
  - dot_ap
  - manhattan_accuracy_threshold
  - manhattan_f1
  - manhattan_f1_threshold
  - manhattan_precision
  - manhattan_recall
  - manhattan_ap
  - euclidean_accuracy_threshold
  - euclidean_f1
  - euclidean_f1_threshold
  - euclidean_precision
  - euclidean_recall
  - euclidean_ap
  - max_accuracy_threshold
  - max_f1
  - max_f1_threshold
  - max_precision
  - max_recall
  - max_ap
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:21900
  - loss:CachedGISTEmbedLoss
widget:
  - source_sentence: temperature in modesto yesterday
    sentences:
      - >-
        Far-sightedness, also known as long-sightedness and hyperopia, is a
        condition of the eye in which light is focused behind, instead of on,
        the retina. This causes close objects to be blurry, while far objects
        may appear normal. As the condition worsens, objects at all distances
        may be blurred.
      - "Manor (/Ë\x88meɪnÉ\x99r/ MAY-ner) is a city in Travis County, Texas, United States. Manor is located 12 miles northeast of Austin and is part of the Austin-Round Rock metropolitan area. The population was 5,037 at the 2010 census. Manor is one of the faster-growing suburbs of Austin."
      - >-
        Modesto Temperature Yesterday. Maximum temperature yesterday: 101 °F
        (at 3:53 PM) Minimum temperature yesterday: 66 °F (at 6:53 AM) Average
        temperature yesterday: 84 °F. Note: Actual official high and low
        records may vary slightly from our data, if they occured in-between our
        weather recording intervals...
  - source_sentence: >-
      More than 169 countries had reported over 212,000 COVID-19 cases before
      March 19 , 2020 .
    sentences:
      - >-
        As of 23 March , more than 341,000 cases of COVID-19 have been reported
        in 192 countries and territories , resulting in more than 14,700 deaths
        and 99,000 recoveries .
      - >-
        As of 21 March , more than 278,000 cases of COVID-19 have been reported
        in over 186 countries and territories , resulting in more than 11,500
        deaths and 92,000 recoveries.  virus seems to mostly spread between
        people via respiratory droplets .
      - >-
        As of 18 March 2020 , more than 212,000 cases of COVID-19 have been
        reported in at least 170 countries and territories , with major
        outbreaks in China , Iran and the European Union .
  - source_sentence: |-
      Premiership
      Aberdeen 2-0 Partick Thistle
      Hamilton Academical 1-1 Kilmarnock
      Inverness Caledonian Thistle 2-2 Dundee
      Motherwell 0-3 Heart of Midlothian
      Rangers 1-1 Ross County
      Championship
      Dumbarton 2-2 St Mirren
      Dundee United 3-0 Raith Rovers
      Falkirk 2-0 Dunfermline Athletic
      Hibernian 1-1 Ayr United
      Queen of the South 3-0 Greenock Morton
      St Johnstone 2-5 Celtic
    sentences:
      - >-
        Match reports from the weekend Scottish Premiership and Championship
        matches.
      - >-
        The car insurance market is "dysfunctional" and does not reward loyal
        customers, said the chief executive of Aviva, Mark Wilson.
      - >-
        A father who died after being assaulted outside a nightclub in Gwynedd
        has been named as Henry Ayabowei.
  - source_sentence: >-
      Electrical energy can be converted into kinetic energy and heat energy by
      an electric motor.
    sentences:
      - >-
        Solution is the term for a homogeneous mixture of two or more
        substances.
      - >-
        Solution is the term for a homogeneous mixture of two or more
        substances.
      - Electric motors transform electrical energy into kinetic energy.
  - source_sentence: when is season 2 of the ranch coming to netflix
    sentences:
      - >-
        The Ranch (TV series) All episodes are named after American country
        music songs, predominantly Kenny Chesney in part one, George Strait in
        part two, Tim McGraw in part three, and Garth Brooks in part four:[5]
        the first ten episodes premiered on April 1, 2016,[6][7] the second
        batch of ten episodes premiered on October 7, 2016. In April 2016,
        Netflix renewed The Ranch for a second season of 20 episodes,[8][9] the
        first half of which premiered on June 16, 2017,[10] and the second half
        was released on December 15, 2017.[11]
      - >-
        The Presidential Agent series The Presidential Agent series was written
        by military author, W. E. B. Griffin. The series consists so far of
        eight novels, By Order of the President, The Hostage, The Hunters, The
        Shooters, Black Ops, The Outlaws, Covert Warriors, and Hazardous Duty.
        Like the rest of his novels, Griffin uses military time, along with the
        address of the place, and the chapter titles are never started on a
        separate page. The series is the author's latest.
      - >-
        Ashley Peacock In 1999, Ashley and Maxine get back together, and finally
        marry in September. Ashley also finds out that his uncle, Fred, is
        actually his biological father. Fred tells Ashley about Kathleen and her
        reluctance to be a mother at a young age. Fred also explains to Ashley
        that Beryl, who he believed to be his mother, is actually his aunt, and
        that Fred let her raise Ashley so he could watch him grow up. Ashley
        decides that he wants to meet his birth mother but Fred begs him not to,
        believing it would hurt Beryl. Ashley, however, tracks Kathleen down to
        her home in Oldham. He is initially very bitter towards her for
        abandoning him but they reconcile and Ashley lets Kathleen attend his
        and Maxine's wedding.
model-index:
  - name: SentenceTransformer based on microsoft/deberta-v3-small
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: sts test
          type: sts-test
        metrics:
          - type: pearson_cosine
            value: 0.033928485348000664
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.08944249572062771
            name: Spearman Cosine
          - type: pearson_manhattan
            value: 0.06296467882181725
            name: Pearson Manhattan
          - type: spearman_manhattan
            value: 0.08266825793291849
            name: Spearman Manhattan
          - type: pearson_euclidean
            value: 0.03489200141716902
            name: Pearson Euclidean
          - type: spearman_euclidean
            value: 0.06202473500014035
            name: Spearman Euclidean
          - type: pearson_dot
            value: 0.2554086617921545
            name: Pearson Dot
          - type: spearman_dot
            value: 0.27863958137561534
            name: Spearman Dot
          - type: pearson_max
            value: 0.2554086617921545
            name: Pearson Max
          - type: spearman_max
            value: 0.27863958137561534
            name: Spearman Max
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: NLI v2
          type: NLI-v2
        metrics:
          - type: cosine_accuracy
            value: 1
            name: Cosine Accuracy
          - type: dot_accuracy
            value: 0.125
            name: Dot Accuracy
          - type: manhattan_accuracy
            value: 1
            name: Manhattan Accuracy
          - type: euclidean_accuracy
            value: 1
            name: Euclidean Accuracy
          - type: max_accuracy
            value: 1
            name: Max Accuracy
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: VitaminC
          type: VitaminC
        metrics:
          - type: cosine_accuracy
            value: 0.55078125
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.9503422379493713
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.6542553191489362
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.656802773475647
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.48616600790513836
            name: Cosine Precision
          - type: cosine_recall
            value: 1
            name: Cosine Recall
          - type: cosine_ap
            value: 0.5203148129920425
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.55078125
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 425.30816650390625
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.6542553191489362
            name: Dot F1
          - type: dot_f1_threshold
            value: 262.8174743652344
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.48616600790513836
            name: Dot Precision
          - type: dot_recall
            value: 1
            name: Dot Recall
          - type: dot_ap
            value: 0.5120444819966403
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.5390625
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 107.76934814453125
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.6542553191489362
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 271.5865478515625
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.48616600790513836
            name: Manhattan Precision
          - type: manhattan_recall
            value: 1
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.5208015383309144
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.55078125
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 7.050784111022949
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.6507936507936508
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 17.465972900390625
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.4823529411764706
            name: Euclidean Precision
          - type: euclidean_recall
            value: 1
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.5175301700973289
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.55078125
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 425.30816650390625
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.6542553191489362
            name: Max F1
          - type: max_f1_threshold
            value: 271.5865478515625
            name: Max F1 Threshold
          - type: max_precision
            value: 0.48616600790513836
            name: Max Precision
          - type: max_recall
            value: 1
            name: Max Recall
          - type: max_ap
            value: 0.5208015383309144
            name: Max Ap

SentenceTransformer based on microsoft/deberta-v3-small

This is a sentence-transformers model finetuned from microsoft/deberta-v3-small 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: microsoft/deberta-v3-small
• 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

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-toytest")
# Run inference
sentences = [
    'when is season 2 of the ranch coming to netflix',
    'The Ranch (TV series) All episodes are named after American country music songs, predominantly Kenny Chesney in part one, George Strait in part two, Tim McGraw in part three, and Garth Brooks in part four:[5] the first ten episodes premiered on April 1, 2016,[6][7] the second batch of ten episodes premiered on October 7, 2016. In April 2016, Netflix renewed The Ranch for a second season of 20 episodes,[8][9] the first half of which premiered on June 16, 2017,[10] and the second half was released on December 15, 2017.[11]',
    "Ashley Peacock In 1999, Ashley and Maxine get back together, and finally marry in September. Ashley also finds out that his uncle, Fred, is actually his biological father. Fred tells Ashley about Kathleen and her reluctance to be a mother at a young age. Fred also explains to Ashley that Beryl, who he believed to be his mother, is actually his aunt, and that Fred let her raise Ashley so he could watch him grow up. Ashley decides that he wants to meet his birth mother but Fred begs him not to, believing it would hurt Beryl. Ashley, however, tracks Kathleen down to her home in Oldham. He is initially very bitter towards her for abandoning him but they reconcile and Ashley lets Kathleen attend his and Maxine's wedding.",
]
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]

Evaluation

Metrics

Semantic Similarity

• Dataset: sts-test
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.0339
spearman_cosine	0.0894
pearson_manhattan	0.063
spearman_manhattan	0.0827
pearson_euclidean	0.0349
spearman_euclidean	0.062
pearson_dot	0.2554
spearman_dot	0.2786
pearson_max	0.2554
spearman_max	0.2786

Triplet

• Dataset: NLI-v2
• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	1.0
dot_accuracy	0.125
manhattan_accuracy	1.0
euclidean_accuracy	1.0
max_accuracy	1.0

Binary Classification

• Dataset: VitaminC
• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.5508
cosine_accuracy_threshold	0.9503
cosine_f1	0.6543
cosine_f1_threshold	0.6568
cosine_precision	0.4862
cosine_recall	1.0
cosine_ap	0.5203
dot_accuracy	0.5508
dot_accuracy_threshold	425.3082
dot_f1	0.6543
dot_f1_threshold	262.8175
dot_precision	0.4862
dot_recall	1.0
dot_ap	0.512
manhattan_accuracy	0.5391
manhattan_accuracy_threshold	107.7693
manhattan_f1	0.6543
manhattan_f1_threshold	271.5865
manhattan_precision	0.4862
manhattan_recall	1.0
manhattan_ap	0.5208
euclidean_accuracy	0.5508
euclidean_accuracy_threshold	7.0508
euclidean_f1	0.6508
euclidean_f1_threshold	17.466
euclidean_precision	0.4824
euclidean_recall	1.0
euclidean_ap	0.5175
max_accuracy	0.5508
max_accuracy_threshold	425.3082
max_f1	0.6543
max_f1_threshold	271.5865
max_precision	0.4862
max_recall	1.0
max_ap	0.5208

Training Details

Training Datasets

negation-triplets

• Dataset: negation-triplets
• Size: 1,950 training samples
• Columns: anchor, entailment, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	entailment	negative
  type	string	string	string
  details	min: 5 tokens mean: 22.15 tokens max: 154 tokens	min: 4 tokens mean: 14.26 tokens max: 43 tokens	min: 5 tokens mean: 14.48 tokens max: 43 tokens

• Samples:

  anchor	entailment	negative
  The wound was fatal.	It was a deadly laceration.	It was not a lethal laceration.
  A woman smiles as she cuts up an identification card.	A woman is holding scissors and is going to cut a card	A woman is holding scissors and is not going to cut a card
  In any flea market, keep your own valuables safe pickpockets are not uncommon.	Make sure your safeguard your valuable items when shopping at a flea market.	Make sure you neglect to safeguard your valuable items when shopping at a flea market.

• Loss: CachedGISTEmbedLoss 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: 1,800 training samples
• Columns: claim and evidence
• Approximate statistics based on the first 1000 samples:

  	claim	evidence
  type	string	string
  details	min: 7 tokens mean: 16.25 tokens max: 67 tokens	min: 8 tokens mean: 37.43 tokens max: 224 tokens

• Samples:

  claim	evidence
  There are more than four films in the Bourne film series .	The song '' Extreme Ways '' '' by musician Moby is used as the end title theme of all five films. ''
  The film was rated at 68 % .	Our Idiot Brother received some positive reviews following its release , garnering 68 % approval on Rotten Tomatoes .
  The film Mohenjo Daro got bad views .	The film was released worldwide on 12 August 2016 to generally negative views.

• Loss: CachedGISTEmbedLoss 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: 1,650 training samples
• Columns: sentence2 and sentence1
• Approximate statistics based on the first 1000 samples:

  	sentence2	sentence1
  type	string	string
  details	min: 7 tokens mean: 16.17 tokens max: 41 tokens	min: 7 tokens mean: 14.95 tokens max: 34 tokens

• Samples:

  sentence2	sentence1
  The energy transformations that occur when a candle burns is described by: chemical energy from the wax is converted into light and heat energy.	Which statement best describes the energy transformations that occur when a candle burns?
  Evolution that occurs over a short period of time is known as microevolution.	Evolution that occurs over a short period of time is known as what?
  Children most likely inherit shape of earlobes from their parents.	Which trait do children most likely inherit from their parents?

• Loss: CachedGISTEmbedLoss 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: 1,650 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 7 tokens mean: 23.68 tokens max: 65 tokens	min: 7 tokens mean: 15.16 tokens max: 39 tokens

• Samples:

  sentence1	sentence2
  If students do not know what the process of condensation is, you can tell them it is the opposite of evaporation. In evaporation, a liquid (like water) changes state to become a gas (water vapor). In condensation, a gas (like water vapor) changes state to become a liquid (water).	Evaporation is responsible for changing liquid water into water vapor.
  e. in solids the atoms are closely locked in position and can only vibrate, in liquids the atoms and molecules are more loosely connected and can collide with and move past one another, while in gases the atoms or molecules are free to move independently, colliding frequently.	Within a substance, atoms that collide frequently and move independently of one another are most likely in a gas
  Accordingly, an increase in pressure will cause an increase in density of the gas and a decrease in its volume .	If a gas in a closed area experiences increases in pressure and decreases in temperatures, the volume of the gas will be affected.

• Loss: CachedGISTEmbedLoss 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: 1,800 training samples
• Columns: document and summary
• Approximate statistics based on the first 1000 samples:

  	document	summary
  type	string	string
  details	min: 45 tokens mean: 248.2 tokens max: 439 tokens	min: 9 tokens mean: 25.66 tokens max: 49 tokens

• Samples:

  document	summary
  Brian Martin, 57, and Christopher McMultan, 40, are alleged to have entered Sarah Gloag's home in Perthshire on 19 January. They are accused of holding a knife to her throat, and tying up Mrs Gloag and her husband as well as two children. The men were remanded in custody after appearing in private at Perth Sheriff Court. Sarah Gloag is the step-daughter of Ann Gloag, the founder of the Stagecoach transport company. The charges against Mr Martin and Mr McMultan also allege that they stole jewellery worth £200,000 and £4,000 in cash from the house. Both also face a number of other charges. They made no plea or declaration.	Two men have been accused of abducting members of one of Scotland's richest families at knifepoint.
  Margaret Paterson spent almost £500,000 on designer goods and was found with more than £200,000 in cash in her West End home. Paterson, 61, ran a brothel and escort service from the New Town with business partner Robert Munro, 61, who was also found guilty of the same charge. They will be sentenced on 8 July. Ian Goalen, 59, pleaded guilty to living off the earnings of prostitution at the High Court in Edinburgh on Monday. Goalen then gave evidence against his former bosses. During a nine year period, Paterson and Munro provided prostitutes all over Scotland. Goalen, a former bank manager from East Lothian, acted as a driver for the working girls in Edinburgh, West Lothian, Glasgow, Aberdeen and Newcastle Upon Tyne. However, it came to an end when police raided their premises in Edinburgh's Grosvenor Street in September 2011. Officers found sex toys, designer shoes and evidence which showed Paterson had gone on a £461,604 spending spree in some of Edinburgh's most exclusive shops. Detectives found credit card records which detailed how she bought luxury items from Harvey Nichols, Louis Vuitton and Mulberry. She also purchased health care from the Spire hospital in Murrayfield, Edinburgh. After conviction on Monday, temporary judge Michael O'Grady QC said: "These are serious offences." The trio were convicted of proceeds of crime and immoral earnings charges after a month-long trial at the High Court in Edinburgh. Details can only now be reported as Judge O'Grady passed a contempt of court order at the start of proceedings prohibiting reports of it until the conclusion of the trial. The jury spent three hours deliberating their verdict. They then returned unanimous verdicts on all charges.	An Edinburgh woman who made hundreds of thousands of pounds running a national prostitution racket has been found guilty of living off immoral earnings.
  Since taking charge in 1999, the 61-year-old has led Britain to four successive Olympic team medals, as well as five European team titles. "His knowledge, experience and advice will be sorely missed," said British Equestrian Federation performance director Dan Hughes. The BEF will start the search for a new performance manager later this year. Breisner said: "Having decided after London 2012 that I would step down as eventing performance manager following the Rio 2016 Olympics, now is the time to start the process of appointing my successor. "However, my full concentration and focus remains on our preparations for Rio."	Great Britain eventing team boss Yogi Breisner will step down from his role after this summer's Olympics in Rio.

• Loss: CachedGISTEmbedLoss 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: 1,650 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 7 tokens mean: 17.18 tokens max: 48 tokens	min: 2 tokens mean: 84.6 tokens max: 512 tokens

• Samples:

  sentence1	sentence2
  Molecules in the gas phase can collide with the liquid surface and reenter the liquid via what?	pressure above the liquid. Molecules in the gas phase can collide with the liquid surface and reenter the liquid via condensation. Eventually, a steady state is reached in which the number of molecules evaporating and condensing per unit time is the same, and the system is in a state of dynamic equilibrium. Under these conditions, a liquid exhibits a characteristic equilibrium vapor pressure that depends only on the temperature. We can express the nonlinear relationship between vapor pressure and temperature as a linear relationship using the Clausius–Clapeyron equation. This equation can be used to calculate the enthalpy of vaporization of a liquid from its measured vapor pressure at two or more temperatures. Volatile liquids are liquids with high vapor pressures, which tend to evaporate readily from an open container; nonvolatile liquids have low vapor pressures. When the vapor pressure equals the external pressure, bubbles of vapor form within the liquid, and it boils. The temperature at which a substance boils at a pressure of 1 atm is its normal boiling point.
  What fluid is most prevalent in your body?	Reporting Scientific Work Whether scientific research is basic science or applied science, scientists must share their findings for other researchers to expand and build upon their discoveries. Communication and collaboration within and between sub disciplines of science are key to the advancement of knowledge in science. For this reason, an important aspect of a scientist’s work is disseminating results and communicating with peers. Scientists can share results by presenting them at a scientific meeting or conference, but this approach can reach only the limited few who are present. Instead, most scientists present their results in peer-reviewed articles that are published in scientific journals. Peer-reviewed articles are scientific papers that are reviewed, usually anonymously by a scientist’s colleagues, or peers. These colleagues are qualified individuals, often experts in the same research area, who judge whether or not the scientist’s work is suitable for publication. The process of peer review helps to ensure that the research described in a scientific paper or grant proposal is original, significant, logical, and thorough. Grant proposals, which are requests for research funding, are also subject to peer review. Scientists publish their work so other scientists can reproduce their experiments under similar or different conditions to expand on the findings. The experimental results must be consistent with the findings of other scientists. There are many journals and the popular press that do not use a peer-review system. A large number of online openaccess journals, journals with articles available without cost, are now available many of which use rigorous peer-review systems, but some of which do not. Results of any studies published in these forums without peer review are not reliable and should not form the basis for other scientific work. In one exception, journals may allow a researcher to cite a personal communication from another researcher about unpublished results with the cited author’s permission.
  What is the diffusion of water known as?	Osmosis is the special case of the diffusion of water. It's an important means of transport in cells because the fluid inside and outside cells is mostly water. Water can pass through the cell membrane by simple diffusion, but it can happen more quickly with the help of channel proteins. Water moves in or out of a cell by osmosis until its concentration is the same on both sides of the cell membrane.

• Loss: CachedGISTEmbedLoss 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: 1,650 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 5 tokens mean: 11.41 tokens max: 23 tokens	min: 16 tokens mean: 34.99 tokens max: 62 tokens

• Samples:

  sentence1	sentence2
  what causes acid rain?	the emission of sulfur dioxide causes acid rain. The sulfur dioxide is converted into sulfuric acid.. emissions cause acid rain
  Plant reproduction often requires what?	plant reproduction often requires pollen. Honeybees and other bees transfer the pollen.. Plant reproduction often requires honeybees.
  Water vapor condensing in clouds usually cause what?	water vapor condensing in clouds causes rain. Heavy rain or thunder storms usually cause flash floods.. Water vapor condensing in clouds usually cause flash floods

• Loss: CachedGISTEmbedLoss 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: 1,500 training samples
• Columns: question and fact
• Approximate statistics based on the first 1000 samples:

  	question	fact
  type	string	string
  details	min: 3 tokens mean: 13.8 tokens max: 78 tokens	min: 4 tokens mean: 11.5 tokens max: 30 tokens

• Samples:

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

• Loss: CachedGISTEmbedLoss 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: 1,650 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 4 tokens mean: 8.62 tokens max: 21 tokens	min: 16 tokens mean: 77.34 tokens max: 247 tokens

• Samples:

  sentence1	sentence2
  what is thickening of the sinuses	Sinusitis is an inflammation, thickening, and swelling of the normal tissue called mucosa, which lines all the sinuses.This same type of tissue lines all the passages of your nose as well as the small channels which connect the nose and sinuses.These channels, or ostiomeatal complex, which is pictured above with the gray shading, can become blocked by swollen tissue.his same type of tissue lines all the passages of your nose as well as the small channels which connect the nose and sinuses. These channels, or ostiomeatal complex, which is pictured above with the gray shading, can become blocked by swollen tissue.
  what is the d block element	D-block elements: the transition metals in the middle of the periodic table D-electron: responsible for bonding and reactivity Electron configuration: arrangement of electrons in the orbitals of metals
  how much is airless spray machine	17 The Basics-An Oeriew of Airless Sprayers The Basics-An Oeriew of Airless Sprayers 18 Using a worn tip wastes paint and labor Assume that paint costs $10 per gallon, labor costs $18 an hour, and the contractor sprays 5 gallons of paint per hour.5 The Basics-An Oeriew of Airless Sprayers The Basics-An Oeriew of Airless Sprayers 36 The general purpose of the packings is to create a seal and direct fluid flow. There are two sets of packings, throat and piston: Throat packings seal the displacement rod to the top of the pump cylinder.

• Loss: CachedGISTEmbedLoss 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: 1,650 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 10 tokens mean: 11.85 tokens max: 23 tokens	min: 16 tokens mean: 134.91 tokens max: 512 tokens

• Samples:

  sentence1	sentence2
  where was the movie cahill us marshal filmed	Cahill U.S. Marshal The film was produced by John Wayne's production company Batjac Productions and shot on location in Durango, Mexico.[5]
  where do lymph vessels join the circulatory system	Lymphatic vessel Rhythmic contraction of the vessel walls through movements may also help draw fluid into the smallest lymphatic vessels, capillaries. If tissue fluid builds up the tissue will swell; this is called edema. As the circular path through the body's system continues, the fluid is then transported to progressively larger lymphatic vessels culminating in the right lymphatic duct (for lymph from the right upper body) and the thoracic duct (for the rest of the body); both ducts drain into the circulatory system at the right and left subclavian veins. The system collaborates with white blood cells in lymph nodes to protect the body from being infected by cancer cells, fungi, viruses or bacteria. This is known as a secondary circulatory system.
  who played daniel cleaver in bridget jones diary	Bridget Jones's Diary (film) Bridget Jones's Diary is a 2001 romantic comedy film directed by Sharon Maguire and written by Richard Curtis, Andrew Davies, and Helen Fielding. It is based on Fielding's 1996 novel of the same name, which is a reinterpretation of Jane Austen's Pride and Prejudice. The adaptation stars Renée Zellweger as Bridget, Hugh Grant as the caddish Daniel Cleaver, and Colin Firth as Bridget's "true love", Mark Darcy. Production began in August 2000 and ended in November 2000, and took place largely on location in London and the Home Counties. The film premiered on 4 April 2001 in the United Kingdom and was released to theatres on 13 April 2001 simultaneously in the United Kingdom and in the United States.

• Loss: CachedGISTEmbedLoss 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: 1,500 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 8 tokens mean: 15.52 tokens max: 51 tokens	min: 19 tokens mean: 454.14 tokens max: 512 tokens

• Samples:

  sentence1	sentence2
  Sarah Ferguson became Duchess of where?	Sarah Ferguson - Duchess - Biography.com Sarah Ferguson Duchess of York Sarah Ferguson is the ex-wife of Britain's Prince Andrew and is also a children's book author and film producer. IN THESE GROUPS Sarah Ferguson - Royal Wedding (TV-14; 0:27) An inside look at the royal wedding between Sarah Ferguson and Prince Andrew. Synopsis Born on October 15, 1959 in London, England, Sarah Ferguson married Britain’s Prince Andrew in 1986. The couple divorced ten years later amidst much media tumult. Ferguson has since written children’s books, served as a Weight Watchers representative and done film production work. She has continued to be the object of media scrutiny, having been taped allegedly selling access to her ex-husband. Early Years Duchess Sarah Margaret Ferguson was born on October 15, 1959, in London, England. The second daughter of Major Ronald Ivor Ferguson, Sarah had a privileged English upbringing, attending private boarding school and becoming an accomplished horseback rider. Her father worked as manager of the Prince of Wales' polo team, so Sarah was acquainted with members of the Royal Family from a young age. Her parents divorced when she was 13 and after graduating from secretarial college, Sarah worked for a public relations firm, an art gallery and a publishing company. Duchess of York In 1985, Sarah met Prince Andrew, the Duke of York. The couple married the following year in Westminster Abbey and had two children, Beatrice and Eugenie. Dubbed "Fergie" by the press, Sarah was often criticized for her extravagant lifestyle and outspoken manner. Marriage trouble began to plague the couple, which is often attributed to Prince Andrew's long trips away while serving in the Royal Navy. In 1992, the couple separated, eventually divorcing in 1996 but continuing to live together in separate living quarters. The Duchess of York hosted her own short-lived talk show and appeared in a string of commercials during the 1990s for Weight Watchers. She is the author of an autobiography, some dieting guides and several children's books. Fact Check We strive for accuracy and fairness. If you see something that doesn't look right, contact us ! Citation Information
  Who wrote the words for My Fair Lady and Camelot?	Frederick Loewe Dies at 86 - Wrote 'My Fair Lady' Score - NYTimes.com Frederick Loewe Dies at 86; Wrote 'My Fair Lady' Score By STEPHEN HOLDEN Published: February 15, 1988 Frederick (Fritz) Loewe, the composer who with his longtime lyricist partner Alan Jay Lerner created the scores for ''My Fair Lady,'' ''Camelot,'' ''Paint Your Wagon,'' ''Brigadoon,'' and ''Gigi,'' died yesterday in Palm Springs, Calif.. He was 86 years old. The cause of death was cardiac arrest, according to John F. Morris, an artist and longtime friend of Mr. Loewe. Among the most famous songs Mr. Loewe wrote with Mr. Lerner, who died in June 1986, were ''Almost Like Being in Love,'' ''I Could Have Danced All Night,'' ''On the Street Where You Live,'' ''I've Grown Accustomed to Her Face,'' ''If Ever I Would Leave You,'' ''Gigi,'' and ''Thank Heaven for Little Girls.'' The team's finest songs are marked by a contemporary conversational fluency and precision of phrase joined to a graceful Old World melodicism that looks back often wistfully to the turn-of-the-century operetta. Lerner and Loewe, whose creative chemistry was comparable to that of Rodgers and Hammerstein, Dietz and Schwartz, and George and Ira Gershwin, were a less likely pairing than most other Broadway theatrical teams. Seventeen years older than his New York-born collaborator, Mr. Loewe came from the world of European operetta and in 1924 moved to the United States, where he struggled for years to gain a foothold in the musical theater. The collaboration, which began inauspiciously in 1942, culminated 14 years later with ''My Fair Lady,'' a Broadway show that is widely regarded as the 50's Broadway musical at the pinnacle of perfection. For Rex Harrison, the nonsinging actor who played the role of Henry Higgins, Mr. Loewe invented melodies that matched to perfection his caustic, supercilious delivery. For Julie Andrews, who played the cockney girl whom Higgins turns into a lady, his melodic style extended from the clang of English music hall to the elegance of Straussian waltzes. Soloist With Berlin Symphony Frederick Loewe was born in Berlin on June 10, 1901. His father was Edmund Loewe, a well-known Viennese tenor who created the role of Prince Danilo in ''The Merry Widow'' in 1906. A skillful pianist by the age of 4, Mr. Loewe studied the instrument in Berlin with Ferruccio Busoni and Eugene d'Albert, and worked on composition and orchestration with Emil Nikolaus von Reznicek. At 13, he became the youngest piano soloist to appear with the Berlin Symphony Orchestra. He also began writing songs at a young age, composing the tunes for a music hall sketch in which his father toured Germany. At 15, he wrote ''Katrina,'' a popular song that became an enormous hit across Europe. Mr. Loewe's early songwriting success gave him the confidence to move to the United States in 1924. He gave a concert at Town Hall, followed by an engagement at the Rivoli Theater. But hampered by a tenuous command of the English language and a musical sensibility that was considered not ''American'' enough, he failed to achieve the success he had anticipated. To support himself, he took a succession of odd jobs, from busboy to riding instructor to prizefighter, and even worked out West for several years, as a cowpuncher and a mail carrier. On returning to New York, he played the piano in beer halls and the organ in a movie house, but lost the latter job with the advent of talking pictures. Mr. Loewe married Ernestine Zerline in 1931; they had no children and they were divorced in 1957. To make contact with the musical theater world, he joined the Lambs Club, and in 1935 he finally sold his first song to Broadway. He earned $25 for the tune, ''Love Tiptoed Through My Heart,'' which Dennis King sang in a show called ''Petticoat Fever.'' The following year, another of his songs, ''A Waltz Was Born in Vienna,'' was interpolated into the unsuccessful revue ''The Little Show.'' In 1937, he and Earle Crooker collaborated on a musical, ''Salute to Spring,'' for the St. Louis Opera. And in 1938, he composed his first full
  Where was golf's 1977 US Open held?	1977 US Open Golf Tournament The 1977 U.S. Open was the 77th time the tournament was played. Winner: Hubert Green, 278 Where it was played: Southern Hills Country Club in Tulsa, Oklahoma Tournament dates: June 16-19, 1977 Leader after first round: Larry Nelson, Tom Purtzer, Grier Jones, Florentino Molina, Hubert Green, Rod Funseth and Terry Diehl, 69 Leader after second round: Hubert Green, 136 Leader after third round: Hubert Green, 208 Notable Notes: Hubert Green held a one-stroke lead entering the final round. But after making birdies on holes 12 and 14, Green was informed that a death threat had been made against him by phone, the caller saying he was going to shoot Green on the course. Was it a serious threat? It was taken seriously, anyway: police walked along with Green, and Green walked apart from his fellow competitors. How did Green react? He birdied the two holes immediately after being informed of the threat and won by a stroke. Final Scores

• Loss: CachedGISTEmbedLoss 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: 1,500 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 8 tokens mean: 11.47 tokens max: 20 tokens	min: 14 tokens mean: 57.28 tokens max: 155 tokens

• Samples:

  sentence1	sentence2
  can you see who viewed your post on facebook?	To view the analytics of all your posts at once, click on “Insights” on the menu at the top of the page. Under the “Reach” column, you can see how many people have viewed each of your posts.
  would a brother and sister have the same dna?	Because of recombination, siblings only share about 50 percent of the same DNA, on average, Dennis says. So while biological siblings have the same family tree, their genetic code might be different in at least one of the areas looked at in a given test. That's true even for fraternal twins.
  will government shutdown affect va pensions?	Military Retirees and Survivor Benefit Plan recipients would, during a shutdown, still receive their pension checks as the funding for these benefits is NOT tied to Congress's funding bill. After previous shutdowns, Veteran Affairs lobbied Congress to fund the VA on a two-year budget cycle which exempts the department.

• Loss: CachedGISTEmbedLoss 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: 1,950 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 10 tokens mean: 25.78 tokens max: 56 tokens	min: 10 tokens mean: 25.81 tokens max: 56 tokens

• Samples:

  sentence1	sentence2
  Regionally , Australian cycads are the least at risk , as they are locally low and habitat fragmentation is common .	Regionally , Australian Cycads are the least vulnerable , as they are locally low and habitat fragmentation is common .
  She is currently recording and producing music for other artists and also writing solo tracks under the name Tobora .	She is currently recording and producing music for other artists and writing solo pieces under the name Tobora .
  The original algorithm , however , would divide the new interval into a smaller and a larger subinterval in Step 4 .	However , the original algorithm would divide the new interval in step 4 into a smaller and a larger partial interval .

• Loss: CachedGISTEmbedLoss 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: claim and evidence
• Approximate statistics based on the first 1000 samples:

  	claim	evidence
  type	string	string
  details	min: 9 tokens mean: 21.36 tokens max: 41 tokens	min: 11 tokens mean: 36.11 tokens max: 79 tokens

• Samples:

  claim	evidence
  Dragon Con had over 5000 guests .	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 .
  COVID-19 has reached more than 185 countries .	As of , more than cases of COVID-19 have been reported in more than 190 countries and 200 territories , resulting in more than deaths .
  In March , Italy had 3.6x times more cases of coronavirus than China .	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 .

• Loss: CachedGISTEmbedLoss 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: anchor, entailment, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	entailment	negative
  type	string	string	string
  details	min: 10 tokens mean: 13.84 tokens max: 19 tokens	min: 10 tokens mean: 13.28 tokens max: 21 tokens	min: 10 tokens mean: 13.48 tokens max: 21 tokens

• Samples:

  anchor	entailment	negative
  A very big teddy bear is next to a woman.	A smiling woman at work with a life size teddy bear	An unhappy woman at work with a tiny teddy bear
  Part of train car with a door to the rear connected to the car behind it	A train with a striped door waiting on a train track.	A train with a door without stripes waiting on a train track.
  Full grown cat laying down and sleeping on top of a car.	A cat napping on a red car in a driveway.	A cat not napping on a red car in a driveway.

• Loss: CachedGISTEmbedLoss 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: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 9 tokens mean: 20.81 tokens max: 45 tokens	min: 10 tokens mean: 15.48 tokens max: 23 tokens

• Samples:

  sentence1	sentence2
  humans normally have 23 pairs of chromosomes.	Humans typically have 23 pairs pairs of chromosomes.
  A solution is a homogenous mixture of two or more substances that exist in a single phase.	Solution is the term for a homogeneous mixture of two or more substances.
  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.	Upwelling is the term for when deep ocean water rises to the surface.

• Loss: CachedGISTEmbedLoss 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: document and summary
• Approximate statistics based on the first 1000 samples:

  	document	summary
  type	string	string
  details	min: 61 tokens mean: 245.66 tokens max: 398 tokens	min: 12 tokens mean: 25.88 tokens max: 42 tokens

• Samples:

  document	summary
  Defence Secretary Philip Hammond also said new F-35 Lightning jets would be flying from the base in 2018. After Mr Hammond's briefing Elizabeth Truss, Conservative MP for South West Norfolk, said the development would boost job opportunities. It is also the culmination of a long campaign to keep the base open. Four years ago RAF Marham's future was under threat as plans favoured a transfer of aircraft and facilities to RAF Lossiemouth in Scotland. Ms Truss welcomed the new announcement and said more than 5,000 people were now employed at the base by the RAF and contractors. "Many of these people are highly skilled in disciplines like engineering," she said. "They now have an opportunity to provide maintenance facilities for other countries' aircraft and this will create even more jobs. "Already we know the base is protected until 2040 when the strike fighter goes out of service. "The base is hugely important to the local community as the biggest employer in south west Norfolk with a variety of jobs in many skilled disciplines."	RAF Marham in Norfolk is to become the European maintenance hub for the new generation of strike and fighter aircraft deployed around Europe.
  Police were called to Pier Street, Ventnor, on the Isle of Wight, shortly after midnight to reports a 57-year-old man had been assaulted. Nick Medlin was pronounced dead at the scene. His family said they were "completely devastated". Two men from the island, aged 31 and 32 have been arrested on suspicion of murder. A third man, aged 26, was released on bail pending further inquiries. Mr Medlin, a father-of-two, played bass in a punk band called Manufactured Romance. In a statement released through police, his family said: "We are completely devastated and totally heartbroken by the tragic death of Nick on Christmas Eve. "The family wishes to thank everyone for their kind tributes and ask for privacy to grieve at this very sad time." Hampshire Constabulary has appealed witnesses to come forward. Det Ch Insp Dave Brown said: "Investigations are continuing today as we work to establish the exact circumstances of this man's death. "We would like to speak to anyone who was in or in the vicinity of the Rose Inn on Christmas Eve night." A prison staff blogger, called Know The Danger, posted on Facebook: "Everybody loved Nick Medlin and respected him, and I can say hand on heart he was one of the best officers I have ever worked with in over thirty years. A true professional in every way." A section of Pier Street was cordoned off for much of Christmas Day but has since reopened.	Family and friends have paid tribute to a prison officer killed on a night out in the early hours of Christmas Day.
  The Kirkcaldy side sealed a Premiership play-off quarter-final spot after seven wins from their last 10 games. "The remit this year was just to try and consolidate the club in the division and improve the squad," McKinnon told BBC Radio Scotland. "The play-offs, with the opportunity of the Premiership, is incredible." Rovers will play either Hibernian or Falkirk over two legs in the first stage of the promotion play-offs on 4 and 7 May. Bairns boss Peter Houston has already stated that he is eager to finish second so they avoid the Stark's Park side in that contest. "It's nice to hear them say that because we feel going into these games that we're definitely the underdogs," said McKinnon. "We don't have anything to lose, so anything from now on is a bonus and they should be very competitive games. "I think we're the form team in the league right now, ahead of Rangers. "Anyone coming to play us over two legs is going to find it very difficult. There's more pressure on Hibs and Falkirk to get into the Premiership than there is on us, but that's not to say we won't be giving everything." Raith have not been in Scotland's top flight since 1997 and former Brechin City manager McKinnon is confident his players can handle the challenge ahead. "We've assembled a good squad," he said. "I heard [Hibs boss] Alan Stubbs talking about big players in his dressing room; we're very similar. "We've got winners and their focus now that we're in the play-offs is to try and get in the Premiership. As a team we're going to give it everything we can to try and achieve that."	Raith Rovers manager Ray McKinnon says it would be "incredible" if they returned to the top flight for the first time in almost 20 years.

• Loss: CachedGISTEmbedLoss 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: 128 evaluation samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 7 tokens mean: 17.24 tokens max: 33 tokens	min: 2 tokens mean: 95.92 tokens max: 407 tokens

• Samples:

  sentence1	sentence2
  In the lens makers' equation, diverging lenses and virtual images are associated with what kinds of numbers?	When using the lens makers equation, remember that real things get positive numbers and virtual things get negative numbers. Thus, diverging lenses and virtual images get negative numbers. The object distance is always positive.
  Farsightedness, or hyperopia, is the condition in which distant objects are seen clearly, but nearby objects are?	Farsightedness, or hyperopia, is the condition in which distant objects are seen clearly, but nearby objects are blurry. It occurs when the eyeball is shorter than normal. This causes images to be focused in back of the retina. Hyperopia can be corrected with convex lenses. The lenses focus images farther forward in the eye, so they are on the retina instead of behind it.
  What must replicate in the cell cycle before meiosis i takes place?	Meiosis I begins after DNA replicates during interphase of the cell cycle. In both meiosis I and meiosis II , cells go through the same four phases as mitosis - prophase, metaphase, anaphase and telophase. However, there are important differences between meiosis I and mitosis. The eight stages of meiosis are summarized below. The stages will be described for a human cell, starting with 46 chromosomes.

• Loss: CachedGISTEmbedLoss 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: 128 evaluation samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 6 tokens mean: 11.12 tokens max: 25 tokens	min: 14 tokens mean: 34.98 tokens max: 66 tokens

• Samples:

  sentence1	sentence2
  what are not cells	Viruses are not cells.. Examples include influenza, rabies, HIV, and Herpes viruses.. rabies are not cells
  What are broken down by water?	mechanical weathering is when rocks are broken down by mechanical means. Water is a mechanical weathering force.. rocks are broken down by water
  This process is always more complex in eukaryotes than prokaryotes	Cell division is more complex in eukaryotes than prokaryotes.. Mitosis is cell division.. Eukaryotes have more complex mitosis than prokaryotes.

• Loss: CachedGISTEmbedLoss 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: 128 evaluation samples
• Columns: question and fact
• Approximate statistics based on the first 1000 samples:

  	question	fact
  type	string	string
  details	min: 3 tokens mean: 13.98 tokens max: 47 tokens	min: 4 tokens mean: 11.78 tokens max: 28 tokens

• Samples:

  question	fact
  The thermal production of a stove is generically used for	a stove generates heat for cooking usually
  What creates a valley?	a valley is formed by a river flowing
  when it turns day and night on a planet, what cause this?	a planet rotating causes cycles of day and night on that planet

• Loss: CachedGISTEmbedLoss 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: 128 evaluation samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 4 tokens mean: 8.62 tokens max: 19 tokens	min: 25 tokens mean: 71.45 tokens max: 153 tokens

• Samples:

  sentence1	sentence2
  definition subsidy	A subsidy is a form of financial aid or support extended to an economic sector (or institution, business, or individual) generally with the aim of promoting economic and social policy.
  cost of attending truck driving school	Truck driving school cost can run you anywhere from $1,200 to over $10,000 – depending on the CDL class type, city and state you live in, college or independent school you go to, and how many hours of training. In order to become a truck driver you must have the proper training, as commercial vehicles can be much more difficult to navigate.
  causes of jerking when sleeping	According to the National Institute of Neurological Disorders and Stroke (NINDS), the sudden, involuntary jerking of a muscle or group of muscles while drifting off to sleep is called myoclonus. It is caused by sudden muscle contractions, also known as positive myoclonus, or muscle relaxation, which is referred to as negative myoclonus.

• Loss: CachedGISTEmbedLoss 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: 128 evaluation samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 10 tokens mean: 11.41 tokens max: 18 tokens	min: 24 tokens mean: 132.54 tokens max: 374 tokens

• Samples:

  sentence1	sentence2
  where is the summer palace in st petersburg	Catherine Palace The Catherine Palace (Russian: Екатерининский дворец, Yekaterininskiy dvorets) is a Rococo palace located in the town of Tsarskoye Selo (Pushkin), 30 km south of St. Petersburg, Russia. It was the summer residence of the Russian tsars.
  when did the olsen twins start full house	Mary-Kate and Ashley Olsen In 1987, at the age of six months, the twins were cast in the role of Michelle Tanner on the ABC sitcom Full House. They began filming at nine months old. In order to comply with child labor laws that set strict limits on how long a child actor may work, the sisters took turns playing the role. The Olsens continued to portray Michelle throughout the show's run, which concluded in 1995.
  when did harry potter and the deathly hollows come out	Harry Potter and the Deathly Hallows Harry Potter and the Deathly Hallows is a fantasy novel written by British author J. K. Rowling and the seventh and final novel of the Harry Potter series. The book was released on 21 July 2007, ending the series that began in 1997 with the publication of Harry Potter and the Philosopher's Stone. It was published by Bloomsbury Publishing in the United Kingdom, in the United States by Scholastic, and in Canada by Raincoast Books. The novel chronicles the events directly following Harry Potter and the Half-Blood Prince (2005), and the final confrontation between the wizards Harry Potter and Lord Voldemort.

• Loss: CachedGISTEmbedLoss 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: 119 evaluation samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 8 tokens mean: 15.39 tokens max: 34 tokens	min: 105 tokens mean: 481.03 tokens max: 512 tokens

• Samples:

  sentence1	sentence2
  All children, except one, grow up is the opening line from which famous story?	Peter Pan is the book with the nation's favourite opening line
  Where in an animal would you find a mandible?	Jaw - definition of jaw by The Free Dictionary Jaw - definition of jaw by The Free Dictionary http://www.thefreedictionary.com/jaw n. 1. a. Either of two bony or cartilaginous structures that in most vertebrates form the framework of the mouth and hold the teeth. b. The mandible or maxilla or the part of the face covering these bones. c. Any of various structures of invertebrates that have an analogous function to vertebrate jaws. 2. Either of two opposed hinged parts in a mechanical device. 3. jaws The walls of a pass, canyon, or cavern. 4. jaws A dangerous situation or confrontation: the jaws of death. 5. Slang a. Impudent argument or back talk: Don't give me any jaw. b. A conversation or chat. intr.v. jawed, jaw·ing, jaws Slang 1. To talk vociferously; jabber. 2. To talk; converse. [Middle English jawe, jowe, perhaps from Old French joue, cheek.] jaw′less adj. (dʒɔː) n 1. (Zoology) the part of the skull of a vertebrate that frames the mouth and holds the teeth. In higher vertebrates it consists of the upper jaw (maxilla) fused to the cranium and the lower jaw (mandible). 2. (Zoology) the corresponding part of an invertebrate, esp an insect 3. (Mechanical Engineering) a pair or either of a pair of hinged or sliding components of a machine or tool designed to grip an object 4. slang c. moralizing talk; a lecture vb a. to talk idly; chat; gossip b. to lecture [C14: probably from Old French joue cheek; related to Italian gota cheek] ˈjawˌlike adj (dʒɔ) n. 1. either of two tooth-bearing bones or bony structures, the mandible or maxilla, forming the framework of the vertebrate mouth. 2. the part of the face covering these bones. 3. jaws, anything resembling a pair of jaws in shape or in power to grasp or hold. 4. one of two or more parts, as of a machine, that grasp or hold something or that attach to or mesh with similar parts. 5. Slang. an idle chat. v.i. 6. Slang. to chat; gossip. [1325–75; Middle English jawe, jowe < Old French joue; orig. uncertain] jaw′less, adj. jaw (jô) 1. Either of two bony or cartilaginous structures that in most vertebrate animals form the framework of the mouth, hold the teeth, and are used for biting and chewing food. The lower, movable part of the jaw is called the mandible. The upper, fixed part is called the maxilla. 2. Any of various structures of invertebrate animals, such as the pincers of spiders or mites, that function similarly to the jaws of vertebrates. jaw I will have been jawing you will have been jawing he/she/it will have been jawing we will have been jawing you will have been jawing they will have been jawing Past Perfect Continuous Noun 1. jaw - the part of the skull of a vertebrate that frames the mouth and holds the teeth bone , os - rigid connective tissue that makes up the skeleton of vertebrates maxilla , maxillary , upper jaw , upper jawbone - the jaw in vertebrates that is fused to the cranium alveolar arch - the part of the upper or lower jawbones in which the teeth are set alveolar process , alveolar ridge , gum ridge - a ridge that forms the borders of the upper and lower jaws and contains the sockets of the teeth skull - the bony skeleton of the head of vertebrates chop - a jaw; "I'll hit him on the chops" 2. jaw - the bones of the skull that frame the mouth and serve to open it; the bones that hold the teeth face , human face - the front of the human head from the forehead to the chin and ear to ear; "he washed his face"; "I wish I had seen the look on his face when he got the news" feature , lineament - the characteristic parts of a person's face: eyes and nose and mouth and chin; "an expression of pleasure crossed his features"; "his lineaments were very regular" 3. jaw - holding device consisting of one or both of the opposing parts of a tool that close to hold an object alligator clip , bulldog clip - a clip with a spring that closes the metal jaws chuck - a holding device consisting of adjustable jaws that center a workpiece in a lathe or center a tool in a drill holding device - a device for holding something pair of pliers , pliers , plyers - a gripping hand
  Which team lost the first Super Bowl of the 1980s?	Super Bowl History 1980 - 1989 - Superbowl in the 1980's Super Bowl History 1980 - 1989 Super Bowl XIV Chuck Noll's Pittsburgh Steelers would repeat to win Super Bowl 14 at the Rose Bowl in Pasadena, California on January 20th, 1980 against Ray Malavasi's LA Rams. Terry Bradshaw took home MVP for the second straight year as the Steelers won their 4th Super Bowl before any other team had won three. John Stallworth and Lynn Swan each caught touchdowns, while Franco Harris ran for two. Dave Elmendorf, Rod Perry, and Eddie Brown intercepted three Bradshaw passes, but it wasn't enough. Lawrence McCutcheon connected with Ron Smith on a halfback pass but quarterback Vince Ferragamo couldn't make the big throw for the Rams. Unsung hero, Larry Anderson, had 162 return yards setting up the Steeler win, 31-19. Super Bowl XV Tom Flores' Oakland Raiders beat Dick Vermeil's Philadelphia Eagles, 27-10, in Super Bowl 15 on January 25th, 1981 at the Louisiana Superdome in New Orleans. Ron Jaworski had 291 yards, but was intercepted by linebacker Rod Martin three times. Jim Plunkett threw three touchdowns in Super Bowl Fifteen; an 80 yard bomb to Kenny King, and two shorter scores to Cliff Branch. An Eagle defense led by John Bunting and Herman Edwards couldn't slow Plunkett and Mark Van Eeghen (75 yards). Ted Hendricks, Matt Millen, Dave Browning, and Martin led the stout Raider defense. Super Bowl XVI On January 24, 1982 Super Bowl 16 was played in Pontiac, Michigan at the Pontiac Sliverdome. Bill Walsh's San Francisco 49ers faced Forrest Gregg's Cincinnati Bengals. MVP, Joe Montana, inched his Forty-Niners into Super Bowl Sixteen by completing a last second touchdown to Dwight Clark in the NFC Title Game, known as "The Catch". Montana took home MVP honors, throwing one touchdown to Earl Cooper, while running for another. Ray Wersching had a Super Bowl record 4 field goals. Ken Anderson brought the Bengals roaring back with a touchdown run and pass to Dan Ross. But early turnovers by Chris Collinsworth and Anderson were too much to overcome as Eric Wright, Lynn Thomas, Ronnie Lott, and Dwight Hicks led San Francisco's defense to victory. Super Bowl XVII On January 30th, 1983, Joe Gibbs' Washington Redskins beat Don Shula's Miami Dolphins 27-17 at the Rose Bowl in Pasadena, California. Super Bowl 17 MVP, John Riggins, rushed for a record 166 yards, and Joe Theismann threw two touchdowns, to Alvin Garrett and Charlie Brown, leading the Redskin comeback in the second half. Miami's 17 Super Bowl Seventeen points came in the first half; a 76 yard touchdown pass from David Woodley to Jimmy Cefalo, a short field goal by Uwe Von Schamann, and a 98 yard kickoff return by Fulton Walker. Vernon Dean and Mark Murphy led the Washington defense that held Woodley and Don Strock to 4-17 passing. Super Bowl XVIII Joe Gibbs' Washington Redskins were back as Defending Champs for Super Bowl 18 in Tampa, Florida on January 30th, 1983. Super Bowl Eighteen was different for Joe, as Tom Flores' Los Angeles Raiders blew-out Joe Theismann (2-ints), John Riggins (64-yds) and the rest of the Redskins, 38-9, in the Super Bowl's most lops

• Loss: CachedGISTEmbedLoss 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: 128 evaluation samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 8 tokens mean: 11.7 tokens max: 21 tokens	min: 22 tokens mean: 58.09 tokens max: 108 tokens

• Samples:

  sentence1	sentence2
  what are the admission requirements for university of washington?	University of Washington admissions is selective with an acceptance rate of 49%. Students that get into University of Washington have an average SAT score between 1220-1460 or an average ACT score of 27-32. The regular admissions application deadline for University of Washington is November 15.
  why diwali is called festival of lights?	Learn about India's biggest holiday of the year. Diwali, or Dipawali, is India's biggest and most important holiday of the year. The festival gets its name from the row (avali) of clay lamps (deepa) that Indians light outside their homes to symbolize the inner light that protects from spiritual darkness.
  how connect iphone bluetooth to car?	Connect using Bluetooth Go to Settings > Bluetooth, and turn off Bluetooth. Wait for about 5 seconds, then turn Bluetooth back on. Check the manual that came with your car for more information on how to pair with a Bluetooth device. Most cars require a phone setup on the car display.

• Loss: CachedGISTEmbedLoss 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: 128 evaluation samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 10 tokens mean: 25.72 tokens max: 42 tokens	min: 10 tokens mean: 25.55 tokens max: 41 tokens

• Samples:

  sentence1	sentence2
  They were there to enjoy us and they were there to pray for us .	They were there for us to enjoy and they were there for us to pray .
  After the end of the war in June 1902 , Higgins left Southampton in the `` SSBavarian '' in August , returning to Cape Town the following month .	In August , after the end of the war in June 1902 , Higgins Southampton left the `` SSBavarian '' and returned to Cape Town the following month .
  From the merger of the Four Rivers Council and the Audubon Council , the Shawnee Trails Council was born .	Shawnee Trails Council was formed from the merger of the Four Rivers Council and the Audubon Council .

• Loss: CachedGISTEmbedLoss 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}
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 160
• per_device_eval_batch_size: 64
• gradient_accumulation_steps: 8
• learning_rate: 4e-05
• weight_decay: 0.0001
• num_train_epochs: 0.1
• lr_scheduler_type: cosine_with_min_lr
• lr_scheduler_kwargs: {'num_cycles': 0.5, 'min_lr': 1.3333333333333335e-05}
• warmup_ratio: 0.33
• save_safetensors: False
• fp16: True
• push_to_hub: True
• hub_model_id: bobox/DeBERTa-small-ST-v1-toytest-checkpoints-tmp
• hub_strategy: all_checkpoints
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 160
• per_device_eval_batch_size: 64
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 8
• eval_accumulation_steps: None
• learning_rate: 4e-05
• weight_decay: 0.0001
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 0.1
• max_steps: -1
• lr_scheduler_type: cosine_with_min_lr
• lr_scheduler_kwargs: {'num_cycles': 0.5, 'min_lr': 1.3333333333333335e-05}
• warmup_ratio: 0.33
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: False
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: False
• fp16: True
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: True
• resume_from_checkpoint: None
• hub_model_id: bobox/DeBERTa-small-ST-v1-toytest-checkpoints-tmp
• hub_strategy: all_checkpoints
• hub_private_repo: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• dispatch_batches: None
• split_batches: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	negation-triplets loss	vitaminc-pairs loss	qasc pairs loss	scitail-pairs-pos loss	gooaq pairs loss	xsum-pairs loss	paws-pos loss	nq pairs loss	msmarco pairs loss	openbookqa pairs loss	trivia pairs loss	sciq pairs loss	NLI-v2_max_accuracy	VitaminC_max_ap	sts-test_spearman_cosine
0.0548	1	6.851	5.2593	2.7279	7.9013	1.9180	8.1263	6.3900	2.2178	10.4461	10.6071	4.7477	7.8702	1.1206	1.0	0.5179	0.0705
0.1096	2	7.0772	5.2441	2.6973	6.5699	1.9754	6.6944	6.1687	2.3460	8.0334	7.9983	4.5152	6.7688	0.9838	1.0	0.5208	0.0894

Framework Versions

• Python: 3.10.13
• Sentence Transformers: 3.0.1
• Transformers: 4.42.3
• PyTorch: 2.1.2
• Accelerate: 0.32.1
• Datasets: 2.20.0
• Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}