diff --git "a/checkpoint-2135/README.md" "b/checkpoint-2135/README.md" --- "a/checkpoint-2135/README.md" +++ "b/checkpoint-2135/README.md" @@ -56,98 +56,94 @@ tags: - dataset_size:32500 - loss:GISTEmbedLoss widget: -- source_sentence: uconnect access app cost per month +- source_sentence: Vermont South is bordered by Mitcham to the north , Nunawading + and Forest Hill to the west , Vermont to the south and Wantirna and Ringwood to + the east . sentences: - - "recycling has a positive impact on the environment. Alternatively, through reuse\ - \ and recycling, waste can be converted into useful products. \n Converting waste\ - \ into useful products has a positive impact on the environment." - - The Advantage Package is available now at a new rate of $149.99 a year with over - 15% savings. Uconnect Access is also available at the regular pricing of $14.99 - a month. - - more than 680,000 cases of COVID-19 have been reported in over 190 countries and - territories , resulting in approximately 31,900 deaths . -- source_sentence: who said america is ungovernable + - Vermont South is bordered to the north of Mitcham , to the west by Nunawading + and Forest Hill , to the south by Vermont and to the east by Wantirna and Ringwood + . + - Ocean waves are among the most impressive waves in the world. They clearly show + that waves transfer energy. In the case of ocean waves, energy is transferred + through matter. But some waves, called electromagnetic waves, can transfer energy + without traveling through matter. These waves can travel through space. You can + read more about electromagnetic waves in the chapter "Electromagnetic Radiation. + " Waves that transfer energy through matter are the focus of the present chapter. + These waves are called mechanical waves. + - Mosque–Cathedral of Córdoba The building is most notable for its arcaded hypostyle + hall, with 856 columns of jasper, onyx, marble, and granite. These were made from + pieces of the Roman temple which had occupied the site previously, as well as + other destroyed Roman buildings, such as the Mérida amphitheatre. The double arches + were a new introduction to architecture, permitting higher ceilings than would + otherwise be possible with relatively low columns. The double arches consist of + a lower horseshoe arch and an upper semi-circular arch. The famous alternating + red and white voussoirs of the arches were inspired by those in the Dome of the + Rock.[31] and also resemble those of the Aachen Cathedral, which were built almost + at the same time. Horseshoe arches were known in the Iberian Peninsula since late + Antiquity, as can be seen on the 3rd-century "Estela de los Flavios", now in the + arqueological museum of León. A centrally located honey-combed dome has blue tiles + decorated with stars. +- source_sentence: What are single-celled organisms that lack a nucleus? sentences: - - '1 Variant translations or reports: 2 America is ungovernable; those who served - the revolution have plowed the sea. 3 As quoted in Man, State, and Society in - Latin American History (1972) by Sheldon B. Liss and Peggy K. Liss, p. 133. 4 Those - who have served the cause of the revolution have plowed the sea.' - - a white stove turned off with a digital clock - - As of 19 March , more than 223,000 cases of COVID-19 have been reported in over - 150 countries and territories , resulting in more than 9,100 deaths and 85,000 - recoveries . -- source_sentence: A man in a wheelchair and another sitting on a bench that is overlooking - the water. + - a dog standing near a street sign on a dirt road + - "Prokaryotes are single-celled organisms that lack a nucleus.. Prokaryotes All\ + \ bacteria are prokaryotes. \n bacteria are single-celled organisms that lack\ + \ a nucleus." + - 1 cup rice uncooked = 7 oz / 200 g = 600 g ( 5 cups / 21 oz in weight) cooked + [2] (Will serve 5 people). 1 pound of rice = 2 1/4 to 2 1/2 cups uncooked = 11 + cups cooked (Will serve 11 people). +- source_sentence: In the UK television series ‘On The Buses’, what is the first name + of Inspector Blake? sentences: - - Two people sitting on dock looking at the ocean. - - an landfill is a source of pollution - - A lymphocyte is the type of white blood cell involved in an immune system response. - You can see what a lymphocyte looks like, greatly magnified, in Figure below . - Lymphocytes make up about one quarter of all white blood cells, but there are - trillions of them in the human body. Usually, fewer than half of the body’s lymphocytes - are in the blood. The majority are in the lymph, lymph nodes, and lymph organs. -- source_sentence: what is the normal percentage of neutrophils + - How Long Does Vicodin Stay in Your System? While the effects of Vicodin last around + 4 hours, traces of the drug can remain in your system for up to 72 hours. In some + adults you can still find traces of the drug in your system after 5 days. A person's + weight, diet, age and level of body fat will contribute to how quickly your body + is able to process and rid itself of the drug. In general, the higher your metabolism + rate, the sooner traces of the drug will be eliminated. + - 'Stephen Lewis - IMDb IMDb Actor | Writer Stephen Lewis, will be chiefly remembered + for the comedy catchphrase: "I ''ate you Butler!" He delivered it week after week + in the hit sitcom On The Buses, a saucy slice of life that ran on ITV from 1969 + to 1973. Lewis was Cyril "Blakey" Blake, a bus inspector with a Hitler moustache + and delusions of grandeur. His nemesis was Stan Butler, a driver ... See full + bio » Born:' + - 'Definition of Dysuria Our Symptoms article on Burning Urination provides a comprehensive + look at the possible causes and treatments of Burning Urination. Definition of + Dysuria Dysuria: Pain during urination, or difficulty urinating. Dysuria is usually + caused by inflammation of the urethra, frequently as a result of infection. Last + Editorial Review: 5/13/2016' +- source_sentence: 'Two cups of black coffee sitting next to a coffee pot. ' sentences: - - The two planes are different in design, but flying in a similar flight pattern. - - 'Normal Blood Counts: The normal range for WBC is generally from 4.0 to 11.0 k/ul. - The normal range of the percentage of neutrophils is between 45% and 70%. blood - cells, filter foreign substances from the blood, and produce lymphocytes(white - blood cells). BIOL 230 Lab Manual Appendix D The Complete Blood Count: An increase - in the percentage of neutrophils (neutrophilia) is seen with an infection.' - - The energy of electromagnetic waves depends on their frequency. Low-frequency - waves have little energy and are normally harmless. High-frequency waves have - a lot of energy and are potentially very harmful. -- source_sentence: 'A van drove into crowds of people in the centre of the city, in - a tourist area called Las Ramblas. - - Thirteen people have died and dozens were injured. - - If you''re upset by anything in the news take a look at the advice here. - - Witnesses say the van was deliberately driven into people in Barcelona''s popular - Las Ramblas area. - - The emergency services got to the location very quickly and tried to help as many - people as they could. - - They are now looking for the missing driver. - - Spanish police said they have stopped another van attack in a small town called - Cambrils near Barcelona. - - The police managed to stop the attackers before they reached a crowded area. - - One woman died and the five suspected attackers were killed by police. - - Police think this attack was linked to the one that happened earlier in Barcelona. - - Four people have been arrested. - - The police are still looking for the missing driver who ran away from the van. - - Spain''s Prime Minister Mariano Rajoy has described it as a terrorist attack. - - The extremist group that calls itself Islamic State says it is responsible. - - The city''s famous football club, FC Barcelona, has said it was "deeply saddened" - by the attack. - - Lionel Messi added: "There are many more of us who want to live in a world in - peace, without hate." - - Several famous building all over the world have lit up to show their support for - the people of Spain. - - New York City''s One World Trade centre was lit up in the colours of Spain''s - flag, red and yellow. - - The Eiffel Tower turned off its lights in support of Barcelona.' + - As of 19 March , more than 225,000 cases of COVID-19 have been reported in over + 150 countries and territories , resulting in more than 9,200 deaths and 85,000 + recoveries . + - Other than gametes, normal human cells have a total of 46 chromosomes per cell. + - two small cups filled with coffee next to a silver coffee pot +- source_sentence: Larry Lurex was the original stage name of which late singer? sentences: - - An attack took place in the Spanish city of Barcelona on Thursday 17 August. - - A man in the kitchen putting a bottle in his blazer pocket. - - "lack of moisture in the air causes low amounts of rainfall. Clouds are made of\ - \ moisture and the moisture is from the water evaporating. \n Lack of water in\ - \ the air causes low amounts of rainfall." + - Larry Lurex Story - Freddie Mercury Net Worth Larry Lurex Story Read more... Freddie + Mercury Freddie Mercury Net Worth is $100 Million. Freddie Mercury was born in + Zanzibar and has an estimated net worth of $100 million dollars. As the lead singer + and songwriter with the hugely successful British band, Queen, Freddie Mercury + wrote many hit. Freddie Mercury (born Farrok... Freddie Mercury Net Worth is $100 + Million. Freddie Mercury Net Worth is $100 Million. Freddie Mercury was born in + Zanzibar and has an estimated net worth of $100 million dollars. As the lead singer + and songwriter with the hugely successful British band, Queen, Freddie Mercury + wrote many hit Freddie Mercury , 5 September 1946 - 24 November 1991) was a British + musician, singer and songwriter, best known as the lead vocalist and lyricist + of the rock band Queen. As a performer, he was known for his flamboyant stage + persona and powerful vocals over a four-octave range. As a songwriter, Mercury + composed many hits for Queen, including "Bohemian Rhapsody", "Killer Queen", "Somebody + to Love", "Don't Stop Me Now", "Crazy Little Thing Called Love" and "We Are the + Champions". In addition to his work with Queen, he led a solo career, and also + occasionally served as a producer and guest musician for other artists. He died + of bronchopneumonia brought on by AIDS on 24 November 1991, only one day after + publicly acknowledging he had the disease. Mercury was a Parsi born in Zanzibar + and grew up there and in India until his mid-teens. He has been referred to as + "Britain's first Asian rock star". In 2002, Mercury was placed ... + - The everyday beverage coffee is a commodity second only to oil in worldwide trade. + - more than 680,000 cases of COVID-19 have been reported in over 190 countries and + territories , resulting in approximately 31,900 deaths . model-index: - name: SentenceTransformer based on microsoft/deberta-v3-small results: @@ -159,34 +155,34 @@ model-index: type: sts-test metrics: - type: pearson_cosine - value: 0.6515835819558387 + value: 0.6620027075959773 name: Pearson Cosine - type: spearman_cosine - value: 0.6624880521170792 + value: 0.6712350889173807 name: Spearman Cosine - type: pearson_manhattan - value: 0.6689532703109837 + value: 0.6792900086102078 name: Pearson Manhattan - type: spearman_manhattan - value: 0.6642616788598046 + value: 0.6752350909143282 name: Spearman Manhattan - type: pearson_euclidean - value: 0.6664666041698615 + value: 0.6760063290032935 name: Pearson Euclidean - type: spearman_euclidean - value: 0.6626712190299542 + value: 0.6714677059740932 name: Spearman Euclidean - type: pearson_dot - value: 0.6496631696023635 + value: 0.6601968030872636 name: Pearson Dot - type: spearman_dot - value: 0.660623311806161 + value: 0.6700788953280725 name: Spearman Dot - type: pearson_max - value: 0.6689532703109837 + value: 0.6792900086102078 name: Pearson Max - type: spearman_max - value: 0.6642616788598046 + value: 0.6752350909143282 name: Spearman Max - task: type: binary-classification @@ -196,109 +192,109 @@ model-index: type: allNLI-dev metrics: - type: cosine_accuracy - value: 0.689453125 + value: 0.701171875 name: Cosine Accuracy - type: cosine_accuracy_threshold - value: 0.9275094270706177 + value: 0.9141938090324402 name: Cosine Accuracy Threshold - type: cosine_f1 - value: 0.5508684863523573 + value: 0.5469728601252609 name: Cosine F1 - type: cosine_f1_threshold - value: 0.862393319606781 + value: 0.8089991807937622 name: Cosine F1 Threshold - type: cosine_precision - value: 0.4826086956521739 + value: 0.42810457516339867 name: Cosine Precision - type: cosine_recall - value: 0.6416184971098265 + value: 0.7572254335260116 name: Cosine Recall - type: cosine_ap - value: 0.5163761733021546 + value: 0.5178189449379189 name: Cosine Ap - type: dot_accuracy - value: 0.689453125 + value: 0.701171875 name: Dot Accuracy - type: dot_accuracy_threshold - value: 701.0382080078125 + value: 701.9563598632812 name: Dot Accuracy Threshold - type: dot_f1 - value: 0.5481481481481482 + value: 0.5492662473794548 name: Dot F1 - type: dot_f1_threshold - value: 657.1117553710938 + value: 619.0263671875 name: Dot F1 Threshold - type: dot_precision - value: 0.47844827586206895 + value: 0.4309210526315789 name: Dot Precision - type: dot_recall - value: 0.6416184971098265 + value: 0.7572254335260116 name: Dot Recall - type: dot_ap - value: 0.5170597922539218 + value: 0.5185120023725823 name: Dot Ap - type: manhattan_accuracy - value: 0.689453125 + value: 0.697265625 name: Manhattan Accuracy - type: manhattan_accuracy_threshold - value: 217.35939025878906 + value: 243.82118225097656 name: Manhattan Accuracy Threshold - type: manhattan_f1 - value: 0.5686746987951807 + value: 0.5536480686695279 name: Manhattan F1 - type: manhattan_f1_threshold - value: 311.76495361328125 + value: 353.4998474121094 name: Manhattan F1 Threshold - type: manhattan_precision - value: 0.48760330578512395 + value: 0.4402730375426621 name: Manhattan Precision - type: manhattan_recall - value: 0.6820809248554913 + value: 0.7456647398843931 name: Manhattan Recall - type: manhattan_ap - value: 0.5200494439911988 + value: 0.5208045922745452 name: Manhattan Ap - type: euclidean_accuracy - value: 0.689453125 + value: 0.701171875 name: Euclidean Accuracy - type: euclidean_accuracy_threshold - value: 10.526958465576172 + value: 11.46473503112793 name: Euclidean Accuracy Threshold - type: euclidean_f1 - value: 0.5508684863523573 + value: 0.5458422174840084 name: Euclidean F1 - type: euclidean_f1_threshold - value: 14.494762420654297 + value: 16.86907958984375 name: Euclidean F1 Threshold - type: euclidean_precision - value: 0.4826086956521739 + value: 0.43243243243243246 name: Euclidean Precision - type: euclidean_recall - value: 0.6416184971098265 + value: 0.7398843930635838 name: Euclidean Recall - type: euclidean_ap - value: 0.5163758899855585 + value: 0.5178582923548423 name: Euclidean Ap - type: max_accuracy - value: 0.689453125 + value: 0.701171875 name: Max Accuracy - type: max_accuracy_threshold - value: 701.0382080078125 + value: 701.9563598632812 name: Max Accuracy Threshold - type: max_f1 - value: 0.5686746987951807 + value: 0.5536480686695279 name: Max F1 - type: max_f1_threshold - value: 657.1117553710938 + value: 619.0263671875 name: Max F1 Threshold - type: max_precision - value: 0.48760330578512395 + value: 0.4402730375426621 name: Max Precision - type: max_recall - value: 0.6820809248554913 + value: 0.7572254335260116 name: Max Recall - type: max_ap - value: 0.5200494439911988 + value: 0.5208045922745452 name: Max Ap - task: type: binary-classification @@ -308,109 +304,109 @@ model-index: type: Qnli-dev metrics: - type: cosine_accuracy - value: 0.6640625 + value: 0.658203125 name: Cosine Accuracy - type: cosine_accuracy_threshold - value: 0.8005393743515015 + value: 0.8023818731307983 name: Cosine Accuracy Threshold - type: cosine_f1 - value: 0.6898839137645109 + value: 0.6916802610114192 name: Cosine F1 - type: cosine_f1_threshold - value: 0.7475045323371887 + value: 0.7310751676559448 name: Cosine F1 Threshold - type: cosine_precision - value: 0.5667574931880109 + value: 0.5623342175066313 name: Cosine Precision - type: cosine_recall - value: 0.8813559322033898 + value: 0.8983050847457628 name: Cosine Recall - type: cosine_ap - value: 0.6963810565222734 + value: 0.6886786340579703 name: Cosine Ap - type: dot_accuracy - value: 0.66796875 + value: 0.66015625 name: Dot Accuracy - type: dot_accuracy_threshold - value: 605.2431640625 + value: 615.7371826171875 name: Dot Accuracy Threshold - type: dot_f1 - value: 0.6864686468646866 + value: 0.6916802610114192 name: Dot F1 - type: dot_f1_threshold - value: 567.939208984375 + value: 558.3238525390625 name: Dot F1 Threshold - type: dot_precision - value: 0.5621621621621622 + value: 0.5623342175066313 name: Dot Precision - type: dot_recall - value: 0.8813559322033898 + value: 0.8983050847457628 name: Dot Recall - type: dot_ap - value: 0.6933667039839441 + value: 0.6875993155505726 name: Dot Ap - type: manhattan_accuracy - value: 0.671875 + value: 0.677734375 name: Manhattan Accuracy - type: manhattan_accuracy_threshold - value: 354.006103515625 + value: 363.1766052246094 name: Manhattan Accuracy Threshold - type: manhattan_f1 - value: 0.6949152542372881 + value: 0.6951026856240127 name: Manhattan F1 - type: manhattan_f1_threshold - value: 410.61859130859375 + value: 436.8890075683594 name: Manhattan F1 Threshold - type: manhattan_precision - value: 0.5790960451977402 + value: 0.5541561712846348 name: Manhattan Precision - type: manhattan_recall - value: 0.8686440677966102 + value: 0.9322033898305084 name: Manhattan Recall - type: manhattan_ap - value: 0.7030057331700904 + value: 0.7003574987662966 name: Manhattan Ap - type: euclidean_accuracy - value: 0.666015625 + value: 0.662109375 name: Euclidean Accuracy - type: euclidean_accuracy_threshold - value: 17.45079803466797 + value: 17.46044158935547 name: Euclidean Accuracy Threshold - type: euclidean_f1 - value: 0.6898839137645109 + value: 0.6906710310965629 name: Euclidean F1 - type: euclidean_f1_threshold - value: 19.631027221679688 + value: 20.146350860595703 name: Euclidean F1 Threshold - type: euclidean_precision - value: 0.5667574931880109 + value: 0.5626666666666666 name: Euclidean Precision - type: euclidean_recall - value: 0.8813559322033898 + value: 0.8940677966101694 name: Euclidean Recall - type: euclidean_ap - value: 0.6964584994456302 + value: 0.6887027301873778 name: Euclidean Ap - type: max_accuracy - value: 0.671875 + value: 0.677734375 name: Max Accuracy - type: max_accuracy_threshold - value: 605.2431640625 + value: 615.7371826171875 name: Max Accuracy Threshold - type: max_f1 - value: 0.6949152542372881 + value: 0.6951026856240127 name: Max F1 - type: max_f1_threshold - value: 567.939208984375 + value: 558.3238525390625 name: Max F1 Threshold - type: max_precision - value: 0.5790960451977402 + value: 0.5626666666666666 name: Max Precision - type: max_recall - value: 0.8813559322033898 + value: 0.9322033898305084 name: Max Recall - type: max_ap - value: 0.7030057331700904 + value: 0.7003574987662966 name: Max Ap --- @@ -476,9 +472,9 @@ from sentence_transformers import SentenceTransformer model = SentenceTransformer("bobox/DeBERTa3-s-CustomPoolin-toytest2-step1-checkpoints-tmp") # Run inference sentences = [ - 'A van drove into crowds of people in the centre of the city, in a tourist area called Las Ramblas.\nThirteen people have died and dozens were injured.\nIf you\'re upset by anything in the news take a look at the advice here.\nWitnesses say the van was deliberately driven into people in Barcelona\'s popular Las Ramblas area.\nThe emergency services got to the location very quickly and tried to help as many people as they could.\nThey are now looking for the missing driver.\nSpanish police said they have stopped another van attack in a small town called Cambrils near Barcelona.\nThe police managed to stop the attackers before they reached a crowded area.\nOne woman died and the five suspected attackers were killed by police.\nPolice think this attack was linked to the one that happened earlier in Barcelona.\nFour people have been arrested.\nThe police are still looking for the missing driver who ran away from the van.\nSpain\'s Prime Minister Mariano Rajoy has described it as a terrorist attack.\nThe extremist group that calls itself Islamic State says it is responsible.\nThe city\'s famous football club, FC Barcelona, has said it was "deeply saddened" by the attack.\nLionel Messi added: "There are many more of us who want to live in a world in peace, without hate."\nSeveral famous building all over the world have lit up to show their support for the people of Spain.\nNew York City\'s One World Trade centre was lit up in the colours of Spain\'s flag, red and yellow.\nThe Eiffel Tower turned off its lights in support of Barcelona.', - 'An attack took place in the Spanish city of Barcelona on Thursday 17 August.', - 'lack of moisture in the air causes low amounts of rainfall. Clouds are made of moisture and the moisture is from the water evaporating. \n Lack of water in the air causes low amounts of rainfall.', + 'Larry Lurex was the original stage name of which late singer?', + 'Larry Lurex Story - Freddie Mercury Net Worth Larry Lurex Story Read more... Freddie Mercury Freddie Mercury Net Worth is $100 Million. Freddie Mercury was born in Zanzibar and has an estimated net worth of $100 million dollars. As the lead singer and songwriter with the hugely successful British band, Queen, Freddie Mercury wrote many hit. Freddie Mercury (born Farrok... Freddie Mercury Net Worth is $100 Million. Freddie Mercury Net Worth is $100 Million. Freddie Mercury was born in Zanzibar and has an estimated net worth of $100 million dollars. As the lead singer and songwriter with the hugely successful British band, Queen, Freddie Mercury wrote many hit Freddie Mercury , 5 September 1946 - 24 November 1991) was a British musician, singer and songwriter, best known as the lead vocalist and lyricist of the rock band Queen. As a performer, he was known for his flamboyant stage persona and powerful vocals over a four-octave range. As a songwriter, Mercury composed many hits for Queen, including "Bohemian Rhapsody", "Killer Queen", "Somebody to Love", "Don\'t Stop Me Now", "Crazy Little Thing Called Love" and "We Are the Champions". In addition to his work with Queen, he led a solo career, and also occasionally served as a producer and guest musician for other artists. He died of bronchopneumonia brought on by AIDS on 24 November 1991, only one day after publicly acknowledging he had the disease. Mercury was a Parsi born in Zanzibar and grew up there and in India until his mid-teens. He has been referred to as "Britain\'s first Asian rock star". In 2002, Mercury was placed ...', + 'The everyday beverage coffee is a commodity second only to oil in worldwide trade.', ] embeddings = model.encode(sentences) print(embeddings.shape) @@ -524,100 +520,100 @@ You can finetune this model on your own dataset. | Metric | Value | |:--------------------|:-----------| -| pearson_cosine | 0.6516 | -| **spearman_cosine** | **0.6625** | -| pearson_manhattan | 0.669 | -| spearman_manhattan | 0.6643 | -| pearson_euclidean | 0.6665 | -| spearman_euclidean | 0.6627 | -| pearson_dot | 0.6497 | -| spearman_dot | 0.6606 | -| pearson_max | 0.669 | -| spearman_max | 0.6643 | +| pearson_cosine | 0.662 | +| **spearman_cosine** | **0.6712** | +| pearson_manhattan | 0.6793 | +| spearman_manhattan | 0.6752 | +| pearson_euclidean | 0.676 | +| spearman_euclidean | 0.6715 | +| pearson_dot | 0.6602 | +| spearman_dot | 0.6701 | +| pearson_max | 0.6793 | +| spearman_max | 0.6752 | #### Binary Classification * Dataset: `allNLI-dev` * Evaluated with [BinaryClassificationEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator) -| Metric | Value | -|:-----------------------------|:---------| -| cosine_accuracy | 0.6895 | -| cosine_accuracy_threshold | 0.9275 | -| cosine_f1 | 0.5509 | -| cosine_f1_threshold | 0.8624 | -| cosine_precision | 0.4826 | -| cosine_recall | 0.6416 | -| cosine_ap | 0.5164 | -| dot_accuracy | 0.6895 | -| dot_accuracy_threshold | 701.0382 | -| dot_f1 | 0.5481 | -| dot_f1_threshold | 657.1118 | -| dot_precision | 0.4784 | -| dot_recall | 0.6416 | -| dot_ap | 0.5171 | -| manhattan_accuracy | 0.6895 | -| manhattan_accuracy_threshold | 217.3594 | -| manhattan_f1 | 0.5687 | -| manhattan_f1_threshold | 311.765 | -| manhattan_precision | 0.4876 | -| manhattan_recall | 0.6821 | -| manhattan_ap | 0.52 | -| euclidean_accuracy | 0.6895 | -| euclidean_accuracy_threshold | 10.527 | -| euclidean_f1 | 0.5509 | -| euclidean_f1_threshold | 14.4948 | -| euclidean_precision | 0.4826 | -| euclidean_recall | 0.6416 | -| euclidean_ap | 0.5164 | -| max_accuracy | 0.6895 | -| max_accuracy_threshold | 701.0382 | -| max_f1 | 0.5687 | -| max_f1_threshold | 657.1118 | -| max_precision | 0.4876 | -| max_recall | 0.6821 | -| **max_ap** | **0.52** | +| Metric | Value | +|:-----------------------------|:-----------| +| cosine_accuracy | 0.7012 | +| cosine_accuracy_threshold | 0.9142 | +| cosine_f1 | 0.547 | +| cosine_f1_threshold | 0.809 | +| cosine_precision | 0.4281 | +| cosine_recall | 0.7572 | +| cosine_ap | 0.5178 | +| dot_accuracy | 0.7012 | +| dot_accuracy_threshold | 701.9564 | +| dot_f1 | 0.5493 | +| dot_f1_threshold | 619.0264 | +| dot_precision | 0.4309 | +| dot_recall | 0.7572 | +| dot_ap | 0.5185 | +| manhattan_accuracy | 0.6973 | +| manhattan_accuracy_threshold | 243.8212 | +| manhattan_f1 | 0.5536 | +| manhattan_f1_threshold | 353.4998 | +| manhattan_precision | 0.4403 | +| manhattan_recall | 0.7457 | +| manhattan_ap | 0.5208 | +| euclidean_accuracy | 0.7012 | +| euclidean_accuracy_threshold | 11.4647 | +| euclidean_f1 | 0.5458 | +| euclidean_f1_threshold | 16.8691 | +| euclidean_precision | 0.4324 | +| euclidean_recall | 0.7399 | +| euclidean_ap | 0.5179 | +| max_accuracy | 0.7012 | +| max_accuracy_threshold | 701.9564 | +| max_f1 | 0.5536 | +| max_f1_threshold | 619.0264 | +| max_precision | 0.4403 | +| max_recall | 0.7572 | +| **max_ap** | **0.5208** | #### Binary Classification * Dataset: `Qnli-dev` * Evaluated with [BinaryClassificationEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator) -| Metric | Value | -|:-----------------------------|:----------| -| cosine_accuracy | 0.6641 | -| cosine_accuracy_threshold | 0.8005 | -| cosine_f1 | 0.6899 | -| cosine_f1_threshold | 0.7475 | -| cosine_precision | 0.5668 | -| cosine_recall | 0.8814 | -| cosine_ap | 0.6964 | -| dot_accuracy | 0.668 | -| dot_accuracy_threshold | 605.2432 | -| dot_f1 | 0.6865 | -| dot_f1_threshold | 567.9392 | -| dot_precision | 0.5622 | -| dot_recall | 0.8814 | -| dot_ap | 0.6934 | -| manhattan_accuracy | 0.6719 | -| manhattan_accuracy_threshold | 354.0061 | -| manhattan_f1 | 0.6949 | -| manhattan_f1_threshold | 410.6186 | -| manhattan_precision | 0.5791 | -| manhattan_recall | 0.8686 | -| manhattan_ap | 0.703 | -| euclidean_accuracy | 0.666 | -| euclidean_accuracy_threshold | 17.4508 | -| euclidean_f1 | 0.6899 | -| euclidean_f1_threshold | 19.631 | -| euclidean_precision | 0.5668 | -| euclidean_recall | 0.8814 | -| euclidean_ap | 0.6965 | -| max_accuracy | 0.6719 | -| max_accuracy_threshold | 605.2432 | -| max_f1 | 0.6949 | -| max_f1_threshold | 567.9392 | -| max_precision | 0.5791 | -| max_recall | 0.8814 | -| **max_ap** | **0.703** | +| Metric | Value | +|:-----------------------------|:-----------| +| cosine_accuracy | 0.6582 | +| cosine_accuracy_threshold | 0.8024 | +| cosine_f1 | 0.6917 | +| cosine_f1_threshold | 0.7311 | +| cosine_precision | 0.5623 | +| cosine_recall | 0.8983 | +| cosine_ap | 0.6887 | +| dot_accuracy | 0.6602 | +| dot_accuracy_threshold | 615.7372 | +| dot_f1 | 0.6917 | +| dot_f1_threshold | 558.3239 | +| dot_precision | 0.5623 | +| dot_recall | 0.8983 | +| dot_ap | 0.6876 | +| manhattan_accuracy | 0.6777 | +| manhattan_accuracy_threshold | 363.1766 | +| manhattan_f1 | 0.6951 | +| manhattan_f1_threshold | 436.889 | +| manhattan_precision | 0.5542 | +| manhattan_recall | 0.9322 | +| manhattan_ap | 0.7004 | +| euclidean_accuracy | 0.6621 | +| euclidean_accuracy_threshold | 17.4604 | +| euclidean_f1 | 0.6907 | +| euclidean_f1_threshold | 20.1464 | +| euclidean_precision | 0.5627 | +| euclidean_recall | 0.8941 | +| euclidean_ap | 0.6887 | +| max_accuracy | 0.6777 | +| max_accuracy_threshold | 615.7372 | +| max_f1 | 0.6951 | +| max_f1_threshold | 558.3239 | +| max_precision | 0.5627 | +| max_recall | 0.9322 | +| **max_ap** | **0.7004** |