checkpoint-1884/README.md

metadata

base_model: microsoft/deberta-v2-xlarge
datasets:
  - 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
  - cosine_accuracy_threshold
  - cosine_f1
  - cosine_f1_threshold
  - cosine_precision
  - cosine_recall
  - cosine_ap
  - dot_accuracy
  - dot_accuracy_threshold
  - dot_f1
  - dot_f1_threshold
  - dot_precision
  - dot_recall
  - dot_ap
  - manhattan_accuracy
  - manhattan_accuracy_threshold
  - manhattan_f1
  - manhattan_f1_threshold
  - manhattan_precision
  - manhattan_recall
  - manhattan_ap
  - euclidean_accuracy
  - euclidean_accuracy_threshold
  - euclidean_f1
  - euclidean_f1_threshold
  - euclidean_precision
  - euclidean_recall
  - euclidean_ap
  - max_accuracy
  - 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:99622
  - loss:CachedGISTEmbedLoss
widget:
  - source_sentence: does alcohol cause anxiety
    sentences:
      - "Kim Jong Unâ\x80\x99s Wife â\x80\x98Missingâ\x80\x99, Assumed Pregnant. Kim Jong Unâ\x80\x99s wife, Ri Sol Ju, has reportedly â\x80\x98gone missingâ\x80\x99 after not making any public appearances for the last 40 days, according to data released by North Korea news-monitoring website NK News."
      - "Japan is the worldâ\x80\x99s largest mobile games market, with $6.2 billion in 2015E revenue. Despite having fewer players than China or the U.S., Japan has the highest average mobile games spending of any major country. China has 785 million mobile gamers, 62% of Asiaâ\x80\x99s total."
      - >-
        Alcohol Causes Anxiety and Behavior Changes. Frequent alcohol
        consumption can extremely impair several functions of your brain
        including the area that is responsible for controlling your behavior.
        This is why as you consume alcohol, your behavior, conduct and other
        cognitive abilities will usually be the first to go.
  - source_sentence: >-
      More than 273 people have died from the 2019-20 coronavirus outside
      mainland China .
    sentences:
      - >-
        More than 3,700 people have died : around 3,100 in mainland China and
        around 550 in all other countries combined .
      - >-
        More than 3,200 people have died : almost 3,000 in mainland China and
        around 275 in other countries .
      - more than 4,900 deaths have been attributed to COVID-19 .
  - source_sentence: >-
      The action of flushing the toilet, performed multiple times daily, is the
      single biggest use of water in the home.
    sentences:
      - >-
        What action, performed multiple times daily, is the single biggest use
        of water in the home?
      - >-
        Which of these would most likely improve the air quality in large Texas
        cities?
      - The innermost layer of the sun is known as what?
  - source_sentence: >-
      You call cellular respiration that does not need oxygen to proceed
      anaerobic respiration.
    sentences:
      - A binary molecular compound is made up of two of what?
      - >-
        What do you call cellular respiration that does not need oxygen to
        proceed?
      - Roots grow in length and width from the primary and secondary what?
  - source_sentence: >-
      What was the cause of the disqualification of Swedish pentathlete
      Hans-Gunnar Liljenwall at Mexico City in 1968, the first as a result of
      failing a doping test at an Olympic Games?
    sentences:
      - >-
        'Cogito, ergo sum' - the meaning and origin of this phrase Cogito, ergo
        sum Usually translated from the Latin as 'I think, therefore I am'.
        Origin Possibly the best known of all philosophical quotations; this is
        from the French philosopher René Descartes in Discourse on Method, 1637,
        where he attempted to prove his existence as a thinking being, by
        thinking. 'I think, therefore I am' comes to us in English via two
        translations. Descartes' original statement in French was "Je pense,
        donc je suis". This is such a well-known line that it has spawned
        humorous alternatives, not least: "I'm pink, therefore I'm spam" "René
        Descartes was a drunken fart - I drink therefore I am".
      - "Hans Gunnar Liljenwall - Alchetron, The Free Social Encyclopedia I Love to read n write about Interesting People Hans Gunnar Liljenwall Role\_\_Olympic athlete Born\_\_9 July 1941 (age\_74) (1941-07-09) Hans-Gunnar Liljenwall (born 9 July 1941) is a Swedish modern pentathlete who caused the disqualification of the Swedish men's team at the 1968 Summer Olympics held in Mexico City for his alcohol use. Liljenwall was the first athlete to be disqualified at the Olympics for drug use, following the introduction of anti-doping regulations by the International Olympic Committee in 1967. Sponsored Links Liljenwall reportedly had \"two beers\" to calm his nerves before the pistol shooting portion of the modern pentathlon. The Swedish team eventually had to return their bronze medals."
      - >-
        Erstwhile | Definition of Erstwhile by Merriam-Webster Examples of
        erstwhile in a sentence <there's now a store where erstwhile lay green
        and pleasant pastures> Did You Know? The adverb erstwhile has been part
        of English since the 16th century, but it is formed from two words that
        are much older. It comes from the Old English words ær, meaning "early,"
        and hwīl, which has much the same meaning as the modern word while. (The
        English word ere, meaning "before," is also descendant of ær.) The
        adjective erstwhile, as in erstwhile enemies, joined the language around
        1900. 1569 First Known Use of erstwhile 1569
model-index:
  - name: SentenceTransformer based on microsoft/deberta-v2-xlarge
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: sts test
          type: sts-test
        metrics:
          - type: pearson_cosine
            value: 0.9106921279068343
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.9239059466308052
            name: Spearman Cosine
          - type: pearson_manhattan
            value: 0.9246302028077688
            name: Pearson Manhattan
          - type: spearman_manhattan
            value: 0.9208650089167368
            name: Spearman Manhattan
          - type: pearson_euclidean
            value: 0.9256757227724981
            name: Pearson Euclidean
          - type: spearman_euclidean
            value: 0.9216089091348462
            name: Spearman Euclidean
          - type: pearson_dot
            value: 0.8935209770403933
            name: Pearson Dot
          - type: spearman_dot
            value: 0.8917772707787254
            name: Spearman Dot
          - type: pearson_max
            value: 0.9256757227724981
            name: Pearson Max
          - type: spearman_max
            value: 0.9239059466308052
            name: Spearman Max
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: allNLI dev
          type: allNLI-dev
        metrics:
          - type: cosine_accuracy
            value: 0.732421875
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.8159675598144531
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.6309278350515464
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.6821494102478027
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.49038461538461536
            name: Cosine Precision
          - type: cosine_recall
            value: 0.884393063583815
            name: Cosine Recall
          - type: cosine_ap
            value: 0.5991355834488622
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.736328125
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 879.013916015625
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.6206896551724138
            name: Dot F1
          - type: dot_f1_threshold
            value: 740.831787109375
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.4948453608247423
            name: Dot Precision
          - type: dot_recall
            value: 0.8323699421965318
            name: Dot Recall
          - type: dot_ap
            value: 0.5930218788414645
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.734375
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 642.3578491210938
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.6324786324786326
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 796.259521484375
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.5016949152542373
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.8554913294797688
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.598127538996714
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.728515625
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 20.647186279296875
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.6322869955156951
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 24.933399200439453
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.5164835164835165
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.815028901734104
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.5969327943116731
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.736328125
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 879.013916015625
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.6324786324786326
            name: Max F1
          - type: max_f1_threshold
            value: 796.259521484375
            name: Max F1 Threshold
          - type: max_precision
            value: 0.5164835164835165
            name: Max Precision
          - type: max_recall
            value: 0.884393063583815
            name: Max Recall
          - type: max_ap
            value: 0.5991355834488622
            name: Max Ap
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: Qnli dev
          type: Qnli-dev
        metrics:
          - type: cosine_accuracy
            value: 0.712890625
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.7136298418045044
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.705685618729097
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.6252090930938721
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.5828729281767956
            name: Cosine Precision
          - type: cosine_recall
            value: 0.8940677966101694
            name: Cosine Recall
          - type: cosine_ap
            value: 0.7429006619357093
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.689453125
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 764.2352294921875
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.7067137809187279
            name: Dot F1
          - type: dot_f1_threshold
            value: 678.6840209960938
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.6060606060606061
            name: Dot Precision
          - type: dot_recall
            value: 0.847457627118644
            name: Dot Recall
          - type: dot_ap
            value: 0.7164630756489325
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.71875
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 725.244384765625
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.7078039927404719
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 829.4242553710938
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.6190476190476191
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.826271186440678
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.7439472197082637
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.71875
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 23.21845245361328
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.7014388489208634
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 26.616165161132812
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.609375
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.826271186440678
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.7459968056180196
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.71875
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 764.2352294921875
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.7078039927404719
            name: Max F1
          - type: max_f1_threshold
            value: 829.4242553710938
            name: Max F1 Threshold
          - type: max_precision
            value: 0.6190476190476191
            name: Max Precision
          - type: max_recall
            value: 0.8940677966101694
            name: Max Recall
          - type: max_ap
            value: 0.7459968056180196
            name: Max Ap

SentenceTransformer based on microsoft/deberta-v2-xlarge

This is a sentence-transformers model finetuned from microsoft/deberta-v2-xlarge on the negation-triplets, vitaminc-pairs, scitail-pairs-qa, scitail-pairs-pos, xsum-pairs, sciq_pairs, qasc_pairs, openbookqa_pairs, msmarco_pairs, nq_pairs, trivia_pairs, gooaq_pairs, paws-pos and global_dataset datasets. It maps sentences & paragraphs to a 1536-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: microsoft/deberta-v2-xlarge
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1536 tokens
• Similarity Function: Cosine Similarity
• Training Datasets:
  • negation-triplets
  • vitaminc-pairs
  • scitail-pairs-qa
  • scitail-pairs-pos
  • xsum-pairs
  • sciq_pairs
  • qasc_pairs
  • openbookqa_pairs
  • msmarco_pairs
  • nq_pairs
  • trivia_pairs
  • gooaq_pairs
  • paws-pos
  • global_dataset
• Language: en

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: DebertaV2Model 
  (1): Pooling({'word_embedding_dimension': 1536, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("bobox/DeBERTa2-0.9B-ST-v1-checkpoints-tmp")
# Run inference
sentences = [
    'What was the cause of the disqualification of Swedish pentathlete Hans-Gunnar Liljenwall at Mexico City in 1968, the first as a result of failing a doping test at an Olympic Games?',
    'Hans Gunnar Liljenwall - Alchetron, The Free Social Encyclopedia I Love to read n write about Interesting People Hans Gunnar Liljenwall Role\xa0\xa0Olympic athlete Born\xa0\xa09 July 1941 (age\xa074) (1941-07-09) Hans-Gunnar Liljenwall (born 9 July 1941) is a Swedish modern pentathlete who caused the disqualification of the Swedish men\'s team at the 1968 Summer Olympics held in Mexico City for his alcohol use. Liljenwall was the first athlete to be disqualified at the Olympics for drug use, following the introduction of anti-doping regulations by the International Olympic Committee in 1967. Sponsored Links Liljenwall reportedly had "two beers" to calm his nerves before the pistol shooting portion of the modern pentathlon. The Swedish team eventually had to return their bronze medals.',
    'Erstwhile | Definition of Erstwhile by Merriam-Webster Examples of erstwhile in a sentence <there\'s now a store where erstwhile lay green and pleasant pastures> Did You Know? The adverb erstwhile has been part of English since the 16th century, but it is formed from two words that are much older. It comes from the Old English words ær, meaning "early," and hwīl, which has much the same meaning as the modern word while. (The English word ere, meaning "before," is also descendant of ær.) The adjective erstwhile, as in erstwhile enemies, joined the language around 1900. 1569 First Known Use of erstwhile 1569',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1536]

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

Evaluation

Metrics

Semantic Similarity

• Dataset: sts-test
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.9107
spearman_cosine	0.9239
pearson_manhattan	0.9246
spearman_manhattan	0.9209
pearson_euclidean	0.9257
spearman_euclidean	0.9216
pearson_dot	0.8935
spearman_dot	0.8918
pearson_max	0.9257
spearman_max	0.9239

Binary Classification

• Dataset: allNLI-dev
• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.7324
cosine_accuracy_threshold	0.816
cosine_f1	0.6309
cosine_f1_threshold	0.6821
cosine_precision	0.4904
cosine_recall	0.8844
cosine_ap	0.5991
dot_accuracy	0.7363
dot_accuracy_threshold	879.0139
dot_f1	0.6207
dot_f1_threshold	740.8318
dot_precision	0.4948
dot_recall	0.8324
dot_ap	0.593
manhattan_accuracy	0.7344
manhattan_accuracy_threshold	642.3578
manhattan_f1	0.6325
manhattan_f1_threshold	796.2595
manhattan_precision	0.5017
manhattan_recall	0.8555
manhattan_ap	0.5981
euclidean_accuracy	0.7285
euclidean_accuracy_threshold	20.6472
euclidean_f1	0.6323
euclidean_f1_threshold	24.9334
euclidean_precision	0.5165
euclidean_recall	0.815
euclidean_ap	0.5969
max_accuracy	0.7363
max_accuracy_threshold	879.0139
max_f1	0.6325
max_f1_threshold	796.2595
max_precision	0.5165
max_recall	0.8844
max_ap	0.5991

Binary Classification

• Dataset: Qnli-dev
• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.7129
cosine_accuracy_threshold	0.7136
cosine_f1	0.7057
cosine_f1_threshold	0.6252
cosine_precision	0.5829
cosine_recall	0.8941
cosine_ap	0.7429
dot_accuracy	0.6895
dot_accuracy_threshold	764.2352
dot_f1	0.7067
dot_f1_threshold	678.684
dot_precision	0.6061
dot_recall	0.8475
dot_ap	0.7165
manhattan_accuracy	0.7188
manhattan_accuracy_threshold	725.2444
manhattan_f1	0.7078
manhattan_f1_threshold	829.4243
manhattan_precision	0.619
manhattan_recall	0.8263
manhattan_ap	0.7439
euclidean_accuracy	0.7188
euclidean_accuracy_threshold	23.2185
euclidean_f1	0.7014
euclidean_f1_threshold	26.6162
euclidean_precision	0.6094
euclidean_recall	0.8263
euclidean_ap	0.746
max_accuracy	0.7188
max_accuracy_threshold	764.2352
max_f1	0.7078
max_f1_threshold	829.4243
max_precision	0.619
max_recall	0.8941
max_ap	0.746

Training Details

Training Datasets

negation-triplets

• Dataset: negation-triplets
• Size: 5,025 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: 22.27 tokens max: 91 tokens	min: 5 tokens mean: 13.77 tokens max: 42 tokens	min: 5 tokens mean: 14.08 tokens max: 42 tokens

• Samples:

  anchor	entailment	negative
  A white kitty cat sitting on a bike seat.	a cat on top of a bike parked indoors	a cat underneath a bike parked indoors
  A bathroom with a sink and toilet and a bowl on the counter.	A glass sink that is under a faucet.	A glass sink that is not under a faucet.
  Seven people are jumping in the air, along the shore.	People at the shore are jumping in the air.	People at the shore are not jumping in the air.

• 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': 0.025}
  

vitaminc-pairs

• Dataset: vitaminc-pairs at be6febb
• Size: 5,025 training samples
• Columns: claim and evidence
• Approximate statistics based on the first 1000 samples:

  	claim	evidence
  type	string	string
  details	min: 6 tokens mean: 16.21 tokens max: 45 tokens	min: 9 tokens mean: 37.01 tokens max: 187 tokens

• Samples:

  claim	evidence
  Boss Key Productions is yet to release the BlueStreak video game .	BlueStreak is an upcoming video game developed by Boss Key Productions and published by Nexon .
  Jay-Z appeared on Blue Ivy 's first two albums .	Jay-Z appeared on Blue Ivy 's first two albums as well , and the two frequently collaborated .
  The film was reviewed by more than 140 critics .	On review aggregator Rotten Tomatoes , the film has an approval rating of 80 % based on 142 reviews , with an average rating of 7.1/10 .

• 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': 0.025}
  

scitail-pairs-qa

• Dataset: scitail-pairs-qa at 0cc4353
• Size: 5,025 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 7 tokens mean: 15.97 tokens max: 41 tokens	min: 6 tokens mean: 15.15 tokens max: 33 tokens

• Samples:

  sentence1	sentence2
  The cytoskeleton is the skeleton of the cell.	What is the skeleton of the cell?
  Muscular dystrophy is a a wasting disease.	What type of disease is muscular dystrophy?
  A pumpkin is a fruit.	Which food is a fruit?

• 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': 0.025}
  

scitail-pairs-pos

• Dataset: scitail-pairs-pos at 0cc4353
• Size: 5,025 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 7 tokens mean: 23.62 tokens max: 68 tokens	min: 7 tokens mean: 15.74 tokens max: 40 tokens

• Samples:

  sentence1	sentence2
  There have been five mass extinctions in Earth's history.	Five mass distinctions have radically altered the history of life.
  The ultimate source of energy for life on Earth is the sun.	Ultimately, most life forms get their energy from the sun.
  N Neurotransmitter Any one of numerous chemicals in the nervous system that modify or result in the transmission of nerve impulses between synapses.	Like a runner passing a baton, the transmission of nerve impulses between neurons depends on neurotransmitters.

• 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': 0.025}
  

xsum-pairs

• Dataset: xsum-pairs
• Size: 5,025 training samples
• Columns: summary and document
• Approximate statistics based on the first 1000 samples:

  	summary	document
  type	string	string
  details	min: 8 tokens mean: 25.63 tokens max: 47 tokens	min: 45 tokens mean: 211.46 tokens max: 371 tokens

• Samples:

  summary	document
  Rising British star Hannah Barnes will ride for the newly-formed Canyon//SRAM team in next year's inaugural UCI Women's WorldTour circuit.	Barnes, who has spent two seasons with American team UHC-Healthcare, won the white jersey for the best young rider at this year's Women's Tour. "It's been really good in America but I'm happy to come back to Europe," said the 22-year-old from Northamptonshire. "My whole goal for 2015 was to get some good results and then come back here." Joining Barnes on the team are the overall winner of the Women's Tour, German time trial star Lisa Brennauer, and RideLondon Grand Prix winner, Italy's Barbara Guarischi. The team will ride the full 17-race Women's WorldTour calendar in 2016. Germany's Canyon will supply the bike frames, with American firm SRAM providing the components. The team's nine riders, from six countries, will wear kit from British company Rapha, which has recently announced it will no longer be supplying elite men's outfit Team Sky after 2016. The team will begin its 2016 season when Tiffany Cromwell rides in the Australian national championships in January, with the first full race being the Ladies Tour of Qatar in February. Helping riders qualify for the Rio Olympics will be a focus for the team, as will the big American races, the Tour of California and Philadelphia Cycling Classic, as well as Britain's Women's Tour and the women's race at the Tour de France, La Course. "My main goal for the year is to make the team for the Olympics but with this injury I don't really know," said Barnes, who broke her ankle in August. "I'd like to go back to California and do the Tour there too."
  A doctor was injured when a prisoner tried to escape during a routine medical appointment at a hospital in North Lanarkshire.	The 31-year-old inmate from HMP Shotts was being escorted by G4S security staff when he tried to get away. A doctor who tried to assist the security staff suffered minor injuries during the incident on Friday. The Scottish Prison Service said an investigation had begun into what happened. A spokesman said: "I can confirm there was an incident involving a prisoner from HMP Shotts at Wishaw General Hospital today. "We will be working with Police Scotland in investigating the full circumstances of the incident." A G4S spokesman said: "During a routine hospital appointment a prisoner attempted to evade custody but was immediately apprehended by G4S staff. "The prompt actions of our officers averted a more serious incident and demonstrates the challenging situations our staff can face while carrying out their duties."
  West Ham winger Michail Antonio has been ruled out for the rest of the season with a "significant injury", manager Slaven Bilic says.	The 27-year-old was injured in the Hammers' 1-0 win over Swansea at London Stadium last weekend. "It's a significant injury and he's out for the season," Bilic confirmed. Antonio, who has scored nine goals for the Hammers this season, was called up by England for the first time in August. "It is a big blow. We know what he has been giving. He is one of our best players," Bilic added. He was again called up for England's World Cup qualifier against Lithuania last month but pulled out of the squad with a hamstring injury and has yet to make his international debut. He joined West Ham from Nottingham Forest in 2015 and signed a new four-year deal with the club last summer.

• 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': 0.025}
  

sciq_pairs

• Dataset: sciq_pairs at 2c94ad3
• Size: 5,025 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 7 tokens mean: 16.81 tokens max: 45 tokens	min: 2 tokens mean: 83.54 tokens max: 512 tokens

• Samples:

  sentence1	sentence2
  What chemical affects the onset of puberty and duration?	Puberty lasts from about ages 12 to 18 years and is controlled by hormones.
  Within the chloroplast, synthesis of what takes place in the fluid inside the inner membrane called the stroma?	Figure 4.17 The chloroplast has an outer membrane, an inner membrane, and membrane structures called thylakoids that are stacked into grana. The space inside the thylakoid membranes is called the thylakoid space. The light harvesting reactions take place in the thylakoid membranes, and the synthesis of sugar takes place in the fluid inside the inner membrane, which is called the stroma. Chloroplasts also have their own genome, which is contained on a single circular chromosome.
  What type of scientist uses earth-orbiting telescopes?	Astronomers use many tools to study things in space. Earth-orbiting telescopes view stars and galaxies from the darkness of space ( Figure below ). They may have optical and radio telescopes to see things that the human eye can't see. Spacecraft travel great distances to send back information on faraway places.

• 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': 0.025}
  

qasc_pairs

• Dataset: qasc_pairs at a34ba20
• Size: 5,025 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 5 tokens mean: 11.37 tokens max: 25 tokens	min: 13 tokens mean: 33.52 tokens max: 64 tokens

• Samples:

  sentence1	sentence2
  What can help decrease emissions?	carpooling decreases the amount of cars used to travel to a place. Cars and gasoline-burning engines are large sources of emissions. Carpooling decreases emissions
  What can antibodies (large Y-shaped proteins) recognize and bind to?	Antibodies are large, Y-shaped proteins that recognize and bind to antigens.. ALL immunogens are antigens. Antibodies are large, Y-shaped proteins that can recognize and bind to immunogens
  What process is needed to support a baby in the womb?	Oxygen is essential for cellular respiration for all aerobic organisms.. Less oxygen for the mother means less oxygen for the baby. Mothers need respiration for their babies.

• 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': 0.025}
  

openbookqa_pairs

• Dataset: openbookqa_pairs
• Size: 3,029 training samples
• Columns: question and fact
• Approximate statistics based on the first 1000 samples:

  	question	fact
  type	string	string
  details	min: 3 tokens mean: 13.89 tokens max: 65 tokens	min: 5 tokens mean: 11.54 tokens max: 31 tokens

• Samples:

  question	fact
  A prickly pear absorbs nutrients from	plants absorb nutrients from soil into themselves through their roots
  When it's summer in the USA it's winter	June is during the winter in the southern hemisphere
  To move electrical energy around a field, a person would use	metal is an electrical energy conductor

• 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': 0.025}
  

msmarco_pairs

• Dataset: msmarco_pairs at 28ff31e
• Size: 5,025 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 4 tokens mean: 8.59 tokens max: 25 tokens	min: 15 tokens mean: 75.16 tokens max: 208 tokens

• Samples:

  sentence1	sentence2
  is switzerland expensive	Hi Im Swiss and I can answer your question. For us Swiss people Switzerland is not that expensive because out salary makes up for it (we have one of the highest salaries per/person worldwide). For foreigners visiting Switzerland many will find it very expensive, from public transportation to food.
  what does dhea sulfate do for women	DHEA, DHEA-S. This test measures the level of dehydroepiandrosterone (DHEA) and dehydroepiandrosterone sulfate (DHEA-S) in your blood. It may also be used to check how well your adrenal glands are working. DHEA is a hormone made by your adrenal glands and to a lesser degree by the ovaries and testes. DHEA is changed into DHEA-S in your adrenal glands and liver. In both men and women, the sex hormones estrogen and testosterone depend on DHEA. DHEA also has a role in the making of insulin growth factor-1 (IGF-1).
  how long do i need to keep tax returns	After filing your late taxes, remember you will keep your tax return three years from the date you file rather than three years from that tax year. For those that have old tax returns on file, don’t be in a rush to shred them. While fall cleaning, organize and file away tax returns.

• 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': 0.025}
  

nq_pairs

• Dataset: nq_pairs at f9e894e
• Size: 5,025 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 10 tokens mean: 11.81 tokens max: 25 tokens	min: 19 tokens mean: 129.07 tokens max: 512 tokens

• Samples:

  sentence1	sentence2
  who owns the rights to call of duty	Call of Duty The Call of Duty games are published and owned by Activision. While Infinity Ward is still a developer, Treyarch and Sledgehammer Games also develop several of the titles with the release of the studios' games alternating with each other. Some games have been developed by Gray Matter Interactive, Nokia, Exakt Entertainment, Spark Unlimited, Amaze Entertainment, n-Space, Aspyr, Rebellion Developments, Ideaworks Game Studio, and nStigate Games. The games use a variety of engines, including the id Tech 3, the Treyarch NGL, and the IW engine.
  who sings i wanna get next to you	I Wanna Get Next to You "I Wanna Get Next to You" is a 1976 soul single written, composed and produced by American songwriter and producer Norman Whitfield, and most famously sung by American R&B band Rose Royce. It is the third official single from the Car Wash soundtrack. The song has also become a staple on oldies radio and on adult contemporary stations.
  when was the birth control pill made available	Birth control In 1909, Richard Richter developed the first intrauterine device made from silkworm gut, which was further developed and marketed in Germany by Ernst Gräfenberg in the late 1920s.[152] In 1951, a chemist, named Carl Djerassi from Mexico City made the hormones in progesterone pills using Mexican yams.[153] Djerassi had chemically created the pill but was not equipped to distribute it to patients. Meanwhile, Gregory Pincus and John Rock with help from the Planned Parenthood Federation of America developed the first birth control pills in the 1950s, such as mestranol/noretynodrel, which became publicly available in the 1960s through the Food and Drug Administration under the name Enovid.[146][154] Medical abortion became an alternative to surgical abortion with the availability of prostaglandin analogs in the 1970s and mifepristone in the 1980s.[155]

• 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': 0.025}
  

trivia_pairs

• Dataset: trivia_pairs at a7c36e3
• Size: 5,025 training samples
• Columns: query and answer
• Approximate statistics based on the first 1000 samples:

  	query	answer
  type	string	string
  details	min: 8 tokens mean: 16.85 tokens max: 55 tokens	min: 15 tokens mean: 202.22 tokens max: 411 tokens

• Samples:

  query	answer
  What famous London building is officially called 1 Canada Square?	One Canada Square, London, England Tourist Information Locals and travelers to connect with About London Borough of Tower Hamlets, London, England 51.505-0.0196 One Canada Square (often incorrectly called Canary Wharf , after its location) is a skyscraper in Canary Wharf, London . It was the tallest building in the United Kingdom from 1990 to 2010, standing at 235 metres (770 ft) above ground level and containing 50 storeys. In late 2010, it was surpassed by The Shard (completed in July 2012) which stands at 309.6 metres (1,016 ft). One Canada Square was designed by principal architect Cesar Pelli, who based the design and shape mainly on the World Financial Center and the Elizabeth Tower. The building is clad with expensive stainless steel. One of the predominant features of the building is the pyramid roof which contains a flashing aircraft warning light, a rare feature for buildings in the United Kingdom. The distinctive pyramid pinnacle is at 240 metres (800 ft) above sea level. One Canada Square is primarily used for offices, though there are some retail units on the lower ground floor. It is a prestigious location for offices and as of January 2013 was 100% let. The building is recognised as a London landmark and it has gained much attention through film, television and other media when its status was the tallest building in the United Kingdom and continues to gain attention. Map
  Who is the patron saint of dentists?	St. Apollonia - Saints & Angels - Catholic Online Saints & Angels Author and Publisher - Catholic Online Facts Take the Saints Trivia Quiz now! St. Apollonia, who died in the year 249, was martyred for not renouncing her faith during the reign of Emperor Philip. The account of the life of St. Apollonia was written by St. Dionysius to Fabian, Bishop of Antioch. Apollonia had all her teeth knocked out after being hit in the face by a Christian persecutor under the reign of Emperor Philip. After she was threatened with fire unless she renounced her faith, Apollonia jumped into the flames voluntarily. She is considered the patron of dental diseases and is often invoked by those with toothaches. Ancient art depicts her with a golden tooth at the end of her necklace. Also in art, she is seen with pincers holding a tooth.
  Which country has the international vehicle registration CDN?	Why does Canada have the letters CDN as its international car registration plate? Surely it would be more logical to have CND?

• 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': 0.025}
  

gooaq_pairs

• Dataset: gooaq_pairs at b089f72
• Size: 5,025 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 8 tokens mean: 11.43 tokens max: 21 tokens	min: 13 tokens mean: 55.2 tokens max: 126 tokens

• Samples:

  sentence1	sentence2
  what are fettuccine used for?	Long flattish noodle-shaped pasta, similar to tagliatelle. A very good pasta to serve with oil or butter-based sauces, as the sauce goes a long way, coats the pasta evenly and also helps to prevent the strands of pasta from clumping together.
  can you drink lemon water after brushing teeth?	DO NOT brush your teeth for at least 30 minutes after drinking the lemon water. Use a soft toothbrush and fluoridated toothpaste (fluoride toughens your enamel) and do not brush aggressively. Acid softens the enamel and makes it more prone to erosion during brushing.
  is ui ux front end?	User experience (UX) design is centered around the satisfaction the user experiences with your software. Front-end development is the technical implementation of the software's user interface (UI). UI design is the graphical bridge that connects the two.

• 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': 0.025}
  

paws-pos

• Dataset: paws-pos at 161ece9
• Size: 5,025 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 10 tokens mean: 25.54 tokens max: 51 tokens	min: 10 tokens mean: 25.51 tokens max: 50 tokens

• Samples:

  sentence1	sentence2
  A railway station of the same name on the Golitsyno -- Minsk railway , is located in Moscow .	A railway station of the same name on the Golitsyno -- Minsk railway line is located in Moscow .
  To calculate such a point mass , an integration is carried out over the entire range of the continuous variable , on the probability density of the random part .	In order to calculate such a point mass , an integration over the entire range of continuous size is carried out on the probability density of the random part .
  It also has representation at the local and regional level .	It also has a representation at regional and local level .

• 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': 0.025}
  

global_dataset

• Dataset: global_dataset
• Size: 36,293 training samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 4 tokens mean: 29.83 tokens max: 328 tokens	min: 2 tokens mean: 52.56 tokens max: 479 tokens

• Samples:

  sentence1	sentence2
  An autotroph ( from Greek autos = self and trophe = nutrition ) is an organism that makes organic compounds from simple molecules .	The word autotroph comes from the Greek autos = self and trophe = nutrition , related to trephein = to make solid , congeal , thicken
  Nonrenewable resources cannot be replaced as easily as it is consumed.	What type of resource cannot be replaced as easily as it is consumed?
  The data included email addresses and passwords that had been stored without any protection, a security firm said. Leaked Source said the massive cache of credentials dated from 2012 but had only now been leaked and put online. And it had come from a hacker who had supplied security firms with 43 million user names from music service Last.fm. Rambler has been described as the Russian equivalent of Yahoo as it offers email services as well as acting as a news and content hub for its users. "We know about that database," said the service in a statement. "It was leaked March 2014 and contained millions of accounts. Right after the accident we forced our users to change their passwords. "We also have forbidden to use the previously used passwords for the same account." Leaked Source broke the news about the breach and said it had verified some of the data with the help of Russian journalists. . Leaked Source said passwords associated with login names had been stored with "no encryption or hashing". Instead, it said, they had been listed in plain text. Analysis of the long list of passwords showed that "asdasd" was the most popular string, used by more than 723,000 people, it said. The second most popular password among the 98 million users was "asdasd123". In June this year, details of more than 100 million users of the Russian VK.com service were shared online. Copies of the long list of login names and passwords was offered online at a price of one bitcoin (£456).	Login names and passwords for more than 98 million users of the Russian Rambler.ru email service have reportedly been stolen and put online.

• 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': 0.025}
  

Evaluation Datasets

vitaminc-pairs

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

  	claim	evidence
  type	string	string
  details	min: 9 tokens mean: 19.71 tokens max: 38 tokens	min: 9 tokens mean: 32.5 tokens max: 78 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': 0.025}
  

negation-triplets

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

  	anchor	entailment	negative
  type	string	string	string
  details	min: 8 tokens mean: 14.11 tokens max: 46 tokens	min: 6 tokens mean: 12.27 tokens max: 18 tokens	min: 6 tokens mean: 12.65 tokens max: 19 tokens

• Samples:

  anchor	entailment	negative
  a bike leaning on a metal fence next to some flowing water.	A bicycle parked next to a flooded river	A bicycle parked far away from a flooded river.
  A woman is painting a mural of a woman's face.	There is a woman painting.	There is no woman painting.
  A woman sitting at a table while holding a pair of scissors.	A woman smiles and holds up a pair of scissors.	A woman frowns and puts down a pair of scissors.

• 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': 0.025}
  

scitail-pairs-pos

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

  	sentence1	sentence2
  type	string	string
  details	min: 9 tokens mean: 20.13 tokens max: 53 tokens	min: 8 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': 0.025}
  

scitail-pairs-qa

• Dataset: scitail-pairs-qa at 0cc4353
• 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: 16.09 tokens max: 41 tokens	min: 9 tokens mean: 14.87 tokens max: 28 tokens

• Samples:

  sentence1	sentence2
  The field of study known as mathematics is called the language of science.	What field of study is called the language of science?
  Roots grow in length and width from the primary and secondary meristem.	Roots grow in length and width from the primary and secondary what?
  Muscle groups are controlled by the motor cortex .	Muscle groups are controlled by what mechanism?

• 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': 0.025}
  

xsum-pairs

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

  	summary	document
  type	string	string
  details	min: 14 tokens mean: 25.09 tokens max: 39 tokens	min: 58 tokens mean: 230.7 tokens max: 342 tokens

• Samples:

  summary	document
  A diamond ring that belonged to former child star Shirley Temple is going up for auction next month at a starting price of $25m (£17.3m), Sotheby's says.	The auction house said the 9.54-carat "Fancy Deep Blue" ring was bought by Temple's father in 1940 for $7,210. Temple, one of Hollywood's most popular stars in the 1930s, died in 2014 at the age of 85. A private buyer bought the ring from her estate and is now putting it up for auction, Sotheby's said. It is estimated to be worth up to $35 million. Temple pursued a career in politics after leaving the entertainment industry, serving as US ambassador to Ghana and Czechoslovakia. A selection of her belongings, including a dress worn at the 1935 Oscars, has previously been auctioned off by her family.
  A cybersecurity researcher working for anti-virus firm Kaspersky Lab in Russia has been arrested.	Ruslan Stoyanov, a member of Kasperksy's investigations team, was arrested in December but news of his apprehension has only just surfaced. He was arrested as part of an investigation into payments he allegedly received from foreign firms. At Kaspersky, Mr Stoyanov helped look into hack attacks and breaches at Russian companies. In a statement, Kaspkersy Lab said the arrest had nothing to do with his work for the security firm. "Ruslan Stoyanov is under investigation for a period predating his employment at Kaspersky Lab," said the company in a statement. Mr Stoyanov joined Kaspersky in 2012. It added: "We do not possess details of the investigation." Prior to working for Kaspersky, Mr Stoyanov was employed at other security firms. From 2000 to 2006 he was a major in the Russian Ministry of Interior's Moscow cybercrime unit. Information about the reasons for the arrest are scant, but one Russian newspaper linked it to a probe into Sergei Mikhailov - a senior official at Russia's FSB intelligence service. Forbes reported that Mr Stoyanov has been arrested under Article 275 of Russia's criminal code which lets prosecutors charge people for treason for "providing financial, technical, advisory, or other assistance" to other countries or non-Russian organisations seen as hostile.
  Scotland's Russell Knox won the Travelers Championship after Jim Furyk became the first player in PGA Tour history to shoot a round of 58.	Knox, 31, closed with a two-under 68 to beat Jerry Kelly by one shot. Daniel Berger went into the final round as leader but carded a four-over-par 74 to end tied for fifth with Furyk, Robert Garrigus and Tyrone van Aswegen. American Furyk, 46, carded 10 birdies and an eagle in his bogey-free round of 12 under par to finish on 11 under. Knox, who also won in Shanghai in November 2015, becomes the fifth player to have multiple wins in the 2015-16 PGA Tour season. The others are the current top three players in the world - Jason Day, Dustin Johnson and Jordan Spieth - and world number eight Adam Scott. The victory lifts the Scot into contention for at least a wildcard place in the European Ryder Cup team for the biennial tournament against the United States, which takes place at the end of September. "It's been an incredible year for me," said Knox. "I keep believing in myself, I tell myself every day that I'm good enough to be up there and win tournaments. It's been an enjoyable ride." Set to move inside the top 20 in the world rankings, he added that winning this tournament would make it difficult for European captain Darren Clarke "not to pick me". Furyk, who was already one of only six men to have recorded a score of 59 for 18 holes, said: "A million and a half rounds played in the history of the PGA Tour and you look at the great names ahead of me. "It's humbling. To stand alone at 58 is really a cool accomplishment."

• 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': 0.025}
  

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: 16.28 tokens max: 37 tokens	min: 2 tokens mean: 72.52 tokens max: 400 tokens

• Samples:

  sentence1	sentence2
  Internal resistance, or (electrical) resistance in general, involves the resistance of the flow of what?	Internal Resistance As noted before, a 12-V truck battery is physically larger, contains more charge and energy, and can deliver a larger current than a 12-V motorcycle battery. Both are lead-acid batteries with identical emf, but, because of its size, the truck battery has a smaller internal resistance r . Internal resistance is the inherent resistance to the flow of current within the source itself. Figure 21.9 is a schematic representation of the two fundamental parts of any voltage source. The emf (represented by a script E in the figure) and internal resistance r are in series. The smaller the internal resistance for a given emf, the more current and the more power the source can supply.
  If a solute is a gas, increasing the temperature will do what?	If a solute is a gas, increasing the temperature decreases its solubility. For example, less carbon dioxide can dissolve in warm ocean water than in cold ocean water.
  What are usually planted in rows with bare soil in between the rows?	The problem doesn’t stop with plowing. Crops are usually planted in rows with bare soil in between the rows. In places where crops grow only during part of the year, the land may be bare for a few months.

• 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': 0.025}
  

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: 11.02 tokens max: 21 tokens	min: 17 tokens mean: 33.07 tokens max: 60 tokens

• Samples:

  sentence1	sentence2
  What kind of organism use cellulose for their cell walls?	Plants use cellulose for their cell walls.. If the plant is green, it is a producer. Producers use cellulose for their cell walls.
  Energy enters what in the form of sunlight or chemical compounds.	Energy enters ecosystems in the form of sunlight or chemical compounds.. Biomes are global ecosystems. Energy enters biomes in the form of sunlight or chemical compounds.
  What does heat and pressure change into natural gas?	heat and pressure change the remains of prehistoric living things into natural gas. Dinosaurs and Other Prehistoric Creatures Dinosaurs are just one group of prehistoric animals. heat and pressure change the remains of dinosaurs into natural gas

• 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': 0.025}
  

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: 13.96 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': 0.025}
  

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: 8.45 tokens max: 20 tokens	min: 29 tokens mean: 72.06 tokens max: 227 tokens

• Samples:

  sentence1	sentence2
  how long is flight from vegas to california	Flying time from Las Vegas, NV to Los Angeles, CA. The total flight duration from Las Vegas, NV to Los Angeles, CA is 41 minutes. This is the average in-air flight time (wheels up to wheels down on the runway) based on actual flights taken over the past year, including routes like LAS to LAX.
  tangible net worth calculation meaning	A measure of the physical worth of a company, which does not include any value derived from intangible assets such as copyrights, patents and intellectual property. Tangible net worth is calculated by taking a firm's total assets and subtracting the value of all liabilities and intangible assets.Next Up.REAKING DOWN 'Tangible Net Worth'. In terms of a consumer, tangible net worth is the sum of all your tangible assets (cash, home, cars, etc) less any liabilities you may have.
  who is father ferdinand	Ferdinand does indeed fall in love with Prospero's daughter Miranda, aided by the magic of Ariel. Prospero does also have a plan for Ferdinand, which is for him to marry his daughter and cement the reconciliation between Prospero, right Duke of Milan and Alonso, Ferdinand's father and King of Milan. It is interesting to note that Ferdinand is presented as more passive than his romantic counterpart, Miranda.

• 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': 0.025}
  

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: 11.92 tokens max: 22 tokens	min: 27 tokens mean: 133.38 tokens max: 416 tokens

• Samples:

  sentence1	sentence2
  who owns the rights to guns and roses	Guns N' Roses Slash was replaced by Nine Inch Nails touring guitarist Robin Finck in January 1997, who signed a two-year contract with the band in August 1997, making him an official member.[130] Finck was originally recommended by Matt Sorum to Rose a year earlier as a possible second guitarist to complement Slash.[122] Slash's departure was followed shortly thereafter by Matt Sorum in April 1997, who was fired by Rose after getting in an argument about Tobias's inclusion in the band.[131] Sorum later stated Tobias was the "Yoko Ono of Guns N' Roses".[122] McKagan was the last of the Appetite lineup to leave, resigning as bassist in August 1997.[132] McKagan had recently become a father and wrote about his decision to leave in his autobiography, stating "Guns had been paying rent on studios for three years now—from 1994 to 1997—and still did not have a single song. The whole operation was so erratic that it didn't seem to fit with my hopes for parenthood, for stability."[132] McKagan was replaced later that year by former Replacements bassist Tommy Stinson.[133] An actual break-up of Guns N' Roses never occurred, as new players were brought in as the old ones left. Rose reportedly purchased the full rights to the Guns N' Roses name in 1997.[129][134] Slash claimed he and bandmates signed over the name in duress, stating "Axl refused to go onstage one night during the Use Your Illusion tour in 1992 unless the band signed away the name rights to the band. Unfortunately, we signed it. I didn't think he'd go on stage otherwise."[135] Rose denied the claim, saying "(it) Never happened, all made up, fallacy and fantasy. Not one single solitary thread of truth to it. Had that been the case I would have been cremated years ago legally, could've cleaned me out for the name and damages. It's called under duress with extenuating circumstances."[135]
  what is red hot chili peppers otherside about	Otherside "Otherside" refers to former band member Hillel Slovak, who died of a heroin overdose on June 25, 1988. The song talks about his struggles from this addiction.[1]
  when do you have to pay death duties	Estate tax in the United States If an asset is left to a spouse or a federally recognized charity, the tax usually does not apply. In addition, a maximum amount, varying year by year, can be given by an individual, before and/or upon their death, without incurring federal gift or estate taxes:[2] $5,340,000 for estates of persons dying in 2014[3] and 2015,[4] $5,450,000 (effectively $10.90 million per married couple) for estates of persons dying in 2016.[5] Because of these exemptions, it is estimated that only the largest 0.2% of estates in the U.S. will pay the tax.[6]

• 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': 0.025}
  

trivia_pairs

• Dataset: trivia_pairs at a7c36e3
• Size: 128 evaluation samples
• Columns: query and answer
• Approximate statistics based on the first 1000 samples:

  	query	answer
  type	string	string
  details	min: 7 tokens mean: 17.31 tokens max: 40 tokens	min: 29 tokens mean: 206.09 tokens max: 385 tokens

• Samples:

  query	answer
  Which musical features the song Sit Down You’re Rockin’ The Boat?	GUYS & DOLLS (Broadway) - "Sit Down, You're Rockin' the Boat" [LIVE @ The 2009 Tony Awards] - YouTube GUYS & DOLLS (Broadway) - "Sit Down, You're Rockin' the Boat" [LIVE @ The 2009 Tony Awards] Want to watch this again later? Sign in to add this video to a playlist. Need to report the video? Sign in to report inappropriate content. Rating is available when the video has been rented. This feature is not available right now. Please try again later. Published on Jun 19, 2013 The cast of the Broadway revival of the musical GUYS & DOLLS, perform the number "Sit Down, You're Rockin' the Boat" live at the 2009 Tony Awards Category
  "John Wayne played the lead role of whom in the 1958 film ""The Conqueror""?"	The movie so toxic it killed John Wayne: the tragedy of The Conqueror The movie so toxic it killed John Wayne: the tragedy of The Conqueror John Wayne and Susan Hayward in The Conqueror Credit: Rex Chris Bell 17 January 2017 • 6:24pm The Duke playing Genghis Khan in yellowface, filming on a nuclear testing ground? Chris Bell on why nobody escaped the fallout from The Conqueror Towards the end of his life, Howard Hughes – the billionaire tycoon, aviator and filmmaker – had become a recluse. Locked in the penthouse suite at his Xanadu Princess Resort hotel in the Bahamas, he refused to bathe, cut his nails or hair, use a toilet or even open the curtains. Instead, he would sit for hours in his darkened bedroom, naked except for a pink hotel napkin, eating nothing but chocolate bars and chicken, surrounded by dozens of Kleenex boxes that he continuously stacked and rearranged.  But another ritual obsession would come dominate his final few months in 1976: two movies, played continually via a projector on the wall, that he watched over and over again. The first was his favourite film , Ice Station Zebra – Rock Hudson’s tense... Premium
  Which country has won the most soccer World Cups?	Brazil may have won the most World Cup titles, but Germany has been the most consistent team

• 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': 0.025}
  

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: 11.65 tokens max: 18 tokens	min: 23 tokens mean: 57.55 tokens max: 129 tokens

• Samples:

  sentence1	sentence2
  do nz residents need a visa to work in australia?	Most NZ citizens can visit, live and work in Australia without applying for a tourist or work visa. NZ permanent residents need to apply for visas to Australia.
  is clep testing worth it?	Money is a valuable resource, but time is even more valuable. CLEP tests can help you get a degree much faster than the traditional college path. The conventional amount of time to get an undergraduate degree is 4 years, though it can take many students as long as six years.
  does someone know if you block them on iphone?	If you block someone, they do not receieve any notification that they have been blocked. The only way for them to know would be for you to tell them. Furthermore, if they send you an iMessage, it will say that it was delivered on their phone, so they won't even know that you're not seeing their message.

• 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': 0.025}
  

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: 25.58 tokens max: 41 tokens	min: 10 tokens mean: 25.4 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': 0.025}
  

global_dataset

• Dataset: global_dataset
• Size: 325 evaluation samples
• Columns: sentence1 and sentence2
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	min: 5 tokens mean: 31.63 tokens max: 324 tokens	min: 2 tokens mean: 58.18 tokens max: 416 tokens

• Samples:

  sentence1	sentence2
  what height should a floating vanity be	If the piece measures 20 inches high and you want a height of 31 inches, you must mount it 11 inches above the floor. If you want the look of a floating sink with the entire vanity raised 12 inches or higher above the floor, you want a vanity of a smaller height.
  Owner Steve Hayes has been actively trying to sell the club since October and four parties were interested back in February. But in a statement the High Wycombe-based outfit said they were seeking fresh interest following stalled talks. Hayes announced his desire to sell up after Wycombe District Council decided not to back plans for a new stadium. Wasps chairman Mark Rigby added: "London Wasps has a long and successful history and is one of the best known brands in the game, and the board is convinced that the right backer exists. "With a great squad in place and the excellent Dai Young at the helm, we believe we are set to make a strong impact next season. "Time is however short and we urgently need a new investor or consortium to back this belief." The statement said the board, after independent advice, could confirm that London Wasps Holdings Limited remained a going concern. Wasps are enduring a torrid domestic season. They have won just two of their last 12 Premiership fixtures and sit second from bottom in the table. In addition Hayes, who also owns League One football club Wycombe Wanderers, is currently The 50-year-old businessman was one of two men arrested in February as part of Operation Tuleta, the investigation running alongside Operation Weeting, which was set up to probe alleged law-breaking at News International. He is currently on bail until June, subject to further enquiries.	An 'imminent takeover' of London Wasps has fallen through, the Premiership strugglers have confirmed.
  can apple cider vinegar cure urinary infection?	The takeaway Apple cider vinegar may have many health benefits, but it's not a cure for UTIs. If you have a UTI, make an appointment with your doctor. A short course of medication should relieve your symptoms within a few days.

• 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': 0.025}
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 128
• learning_rate: 2e-05
• weight_decay: 0.001
• num_train_epochs: 2
• lr_scheduler_type: cosine_with_min_lr
• lr_scheduler_kwargs: {'num_cycles': 0.5, 'min_lr': 5e-06}
• warmup_ratio: 0.3
• save_safetensors: False
• fp16: True
• push_to_hub: True
• hub_model_id: bobox/DeBERTa2-0.9B-ST-v1-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: 64
• per_device_eval_batch_size: 128
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• learning_rate: 2e-05
• weight_decay: 0.001
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 2
• max_steps: -1
• lr_scheduler_type: cosine_with_min_lr
• lr_scheduler_kwargs: {'num_cycles': 0.5, 'min_lr': 5e-06}
• warmup_ratio: 0.3
• 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/DeBERTa2-0.9B-ST-v1-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

Click to expand

Epoch	Step	Training Loss	sciq pairs loss	trivia pairs loss	msmarco pairs loss	scitail-pairs-qa loss	openbookqa pairs loss	nq pairs loss	global dataset loss	vitaminc-pairs loss	scitail-pairs-pos loss	gooaq pairs loss	paws-pos loss	qasc pairs loss	xsum-pairs loss	negation-triplets loss	Qnli-dev_max_ap	allNLI-dev_max_ap	sts-test_spearman_cosine
0.0102	16	7.1882	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0205	32	9.1489	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0307	48	8.805	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0409	64	5.7489	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0512	80	4.7163	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0614	96	3.176	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0716	112	2.034	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0818	128	1.1278	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0921	144	0.7996	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1004	157	-	0.1090	0.4664	0.7974	0.1105	0.7853	1.1498	0.4867	3.8214	0.0780	0.7758	0.0375	0.8442	0.1444	1.2527	0.6672	0.5230	0.8776
0.1023	160	0.54	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1125	176	0.6267	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1228	192	0.401	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1330	208	0.455	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1432	224	0.308	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1535	240	0.2808	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1637	256	0.319	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1739	272	0.3241	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1841	288	0.2181	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1944	304	0.3247	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2008	314	-	0.0483	0.1067	0.1743	0.0014	0.4045	0.1814	0.2370	3.3627	0.0271	0.2038	0.0234	0.1129	0.0382	0.7010	0.7243	0.5918	0.9171
0.2046	320	0.233	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2148	336	0.1946	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2251	352	0.2393	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2353	368	0.1476	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2455	384	0.1976	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2558	400	0.1139	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2660	416	0.1986	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2762	432	0.2405	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2864	448	0.1519	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.2967	464	0.1508	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3012	471	-	0.0414	0.0961	0.0865	0.0004	0.3847	0.1033	0.1735	2.4167	0.0062	0.1280	0.0252	0.0391	0.0329	0.6828	0.7293	0.6023	0.9200
0.3069	480	0.1457	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3171	496	0.1086	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3274	512	0.1412	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3376	528	0.1538	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3478	544	0.1013	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3581	560	0.1007	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3683	576	0.0853	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3785	592	0.0696	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3887	608	0.1468	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3990	624	0.1314	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4015	628	-	0.0295	0.0647	0.0976	0.0001	0.3722	0.0893	0.2361	3.2355	0.0099	0.1432	0.0264	0.0721	0.0172	0.6764	0.7428	0.6036	0.9153
0.4092	640	0.149	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4194	656	0.1402	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4297	672	0.1056	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4399	688	0.0932	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4501	704	0.0534	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4604	720	0.2175	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4706	736	0.1107	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4808	752	0.2301	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.4910	768	0.2317	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5013	784	0.1084	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5019	785	-	0.0324	0.0932	0.1156	0.0001	0.4128	0.1071	0.2526	3.9009	0.0121	0.1237	0.0254	0.0729	0.0098	0.6337	0.7422	0.6029	0.9171
0.5115	800	0.0799	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5217	816	0.1466	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5320	832	0.1208	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5422	848	0.1021	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5524	864	0.1391	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5627	880	0.185	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5729	896	0.1108	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5831	912	0.0926	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5934	928	0.157	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6023	942	-	0.0307	0.0858	0.0844	0.0025	0.3581	0.1235	0.1544	2.8148	0.0050	0.1563	0.0268	0.0460	0.0177	0.5489	0.7193	0.6065	0.9218
0.6036	944	0.126	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6138	960	0.1416	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6240	976	0.157	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6343	992	0.076	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6445	1008	0.0956	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6547	1024	0.1297	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6650	1040	0.1673	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6752	1056	0.0801	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6854	1072	0.1508	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.6957	1088	0.082	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7027	1099	-	0.0361	0.0896	0.0879	0.0001	0.4053	0.0798	0.3097	4.5101	0.0260	0.1373	0.0255	0.0795	0.0101	0.6732	0.7434	0.5941	0.9084
0.7059	1104	0.112	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7161	1120	0.0565	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7263	1136	0.1297	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7366	1152	0.1792	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7468	1168	0.1376	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7570	1184	0.1362	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7673	1200	0.1589	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7775	1216	0.0846	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7877	1232	0.1241	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7980	1248	0.1532	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8031	1256	-	0.0304	0.0667	0.1060	0.0003	0.3666	0.1305	0.2140	3.1243	0.0114	0.2109	0.0277	0.0328	0.0213	0.5495	0.7479	0.5907	0.9194
0.8082	1264	0.0859	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8184	1280	0.0872	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8286	1296	0.0685	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8389	1312	0.0729	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8491	1328	0.0679	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8593	1344	0.0752	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8696	1360	0.1651	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8798	1376	0.0975	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.8900	1392	0.166	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9003	1408	0.079	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9035	1413	-	0.0356	0.0784	0.0609	0.0003	0.4281	0.0720	0.2313	3.4939	0.0141	0.2300	0.0268	0.0522	0.0061	0.5946	0.7379	0.6052	0.9102
0.9105	1424	0.09	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9207	1440	0.0777	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9309	1456	0.1623	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9412	1472	0.08	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9514	1488	0.0628	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9616	1504	0.1695	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9719	1520	0.0715	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9821	1536	0.1493	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9923	1552	0.0431	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0026	1568	0.0549	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0038	1570	-	0.0429	0.0583	0.0681	0.0019	0.4316	0.1454	0.1614	2.8374	0.0053	0.0735	0.0250	0.0282	0.0120	0.5641	0.7139	0.6081	0.9289
1.0128	1584	0.102	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0230	1600	0.0806	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0332	1616	0.0643	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0435	1632	0.2551	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0537	1648	0.1509	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0639	1664	0.0928	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0742	1680	0.1388	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0844	1696	1.2414	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.0946	1712	4.1558	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1042	1727	-	0.0589	0.2021	9.6693	0.6669	3.1183	1.2794	1.6158	3.3692	0.0116	5.4553	0.0264	4.7383	0.0407	0.9119	0.7052	0.5709	0.9119
1.1049	1728	1.8742	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1151	1744	1.7176	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1253	1760	0.3091	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1355	1776	0.3178	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1458	1792	0.173	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1560	1808	0.1028	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1662	1824	0.1533	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1765	1840	0.2395	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1867	1856	0.2036	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.1969	1872	0.2104	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
1.2046	1884	-	0.0323	0.0504	0.0626	0.0001	0.3769	0.0541	0.1662	2.7120	0.0100	0.0981	0.0282	0.0162	0.0219	0.5891	0.7460	0.5991	0.9239

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.42.4
• PyTorch: 2.4.0+cu121
• Accelerate: 0.32.1
• Datasets: 2.21.0
• 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",
}