metadata

language:
  - en
license: apache-2.0
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dense
  - generated_from_trainer
  - dataset_size:1332110
  - loss:CachedMultipleNegativesRankingLoss
base_model: jhu-clsp/ettin-encoder-17m
widget:
  - source_sentence: how to create a network drive in windows 8
    sentences:
      - >-
        - 1 From your Windows 8 Start Menu (the one with all the tiles) click
        Desktop. 2  Right click the desktop and click personalize. 3  Click
        change desktop icons. 4  Check Computer and click apply. 5  If computer
        is already checked then you can ignore this step. 6  Double click
        computer on the desktop.7  Click the map network drive icon. From your
        Windows 8 Start Menu (the one with all the tiles) click Desktop. 2 
        Right click the desktop and click personalize. 3  Click change desktop
        icons. 4  Check Computer and click apply. 5  If computer is already
        checked then you can ignore this step. 6  Double click computer on the
        desktop.
      - >-
        Change from Public to Private Network in Windows 7, 8 and 10 In Windows
        8.1, to change the network profile, we have to go into the PC Settings
        screen. To do that, open the Charms bar and click on Change PC Settings
        at the bottom.or Windows 8, follow the following procedure. First,
        right-click on the network icon in the Windows 8 system tray and click
        on Open Network and Sharing Center. Here you will see the network you
        are connected to and what type of network Windows 8 has identified it
        as.
      - >-
        - Mapping a folder on a network creates a shared space that anyone on
        the network can easily access. You can map a drive on another computer
        on your network to increase the storage you have access to, or to create
        a collaboration space for you and your coworkers. Follow this guide to
        turn any folder into a network drive.
      - >-
        - 1 Log on to the Windows-based server by using an administrative level
        account. 2  Click Start, click Run, type cmd, and then click OK. 3  Type
        the following command, and then press ENTER to share a folder to NFS
        clients and to allow anonymous access: Start Windows Explorer. 2 
        Right-click the folder that you want to share, and then click Sharing.
        3  Click the NFS Sharing tab, and then click Share this folder.
      - >-
        How to Add a Hard Drive to a Network How to Add a Hard Drive to a
        Network. By Dan Gookin from Troubleshooting Your PC For Dummies, 3rd
        Edition. Putting a hard drive up for grabs on the network is done
        primarily by sharing folders on your computer with other computers on
        the network.
      - >-
        Network Drive Network is primarily installed on the central /network/
        storage server in a local network. Typically, users access the network
        drive using local drive access service from their computer. In most
        operating system, network drive is displayed along with the locally
        installed drives.
      - >-
        - Granite has been used for the past 50-60 years as bench-top surfacing
        in kitchens. Prior to that, most stone kitchen bench-tops were marble.
        With the ability to choose from a huge variety of granite colors you are
        assured your countertop will be an original.Granite is the most durable
        of all surfaces used in counter top construction. When it comes to
        kitchen benchtops, normal use of kitchen knives, cutlery, and cookware
        leaves no scratches.ith the ability to choose from a huge variety of
        granite colors you are assured your countertop will be an original.
        Granite is the most durable of all surfaces used in counter top
        construction. When it comes to kitchen benchtops, normal use of kitchen
        knives, cutlery, and cookware leaves no scratches.
      - >-
        Breastfeeding Frequency There are also wet diaper patterns: 1  Day 1 =
        one wet diaper. 2  Day 2 = 2 wet diapers. 3  Day 3 = 3 wet diapers, and
        so on. 4  Day 6 and older = 6 to 8 wet diapers in a 24-hour period.
        (After your milk increases, the urine will soak the diaper.) 5  Your
        baby's urine is pale in color. 6  You can hear your baby swallowing the
        milk.
  - source_sentence: what happens when something melts
    sentences:
      - >-
        - Melting and Freezing. If solid matters gain enough heat they change
        state solid to liquid. Heat is a form of energy and in this situation it
        is used for the break the bonds of the atoms and molecules. Heated atoms
        and molecules vibrate more quickly and break their bonds. We call this
        process melting changing state solid to liquid.
      - >-
        - The shape of the ice cream will change the melting time depending how
        it changes the ratio of the area of the surface to the volume of the ice
        cream. The more ice cream that is in immediate contact with air, the
        faster it will melt. A lower surface area to volume ratio speeds melting
        because the air can share its heat energy with more of the ice cream at
        one time. Imagine how much easier it is to for an ice cream flake to
        melt than a ball of ice cream.
      - >-
        Kid Rock - Fat Bottom Girls chords G C Oh and give it all you got G D C
        Fat bottomed girls You make the rockin ' world go round G D G Fat
        bottomed girls You make the rockin ' world go round Hey listen here G
        Now your mortgages and homes , I got stiffness in the bones D Ain't no
        beauty queens in this locality ( I tell you ) G Oh but I still get my
        pleasure C Still got my greatest ...
      - >-
        - Ice acts very similar to Snow in the sense that it melts when it comes
        into contact with Fire, is near a Torch, or is touched by Lava for a
        short period of time. Ice is not affected by Fire, but when melted with
        Lava, it will produce Water. This will also happen if Ice is destroyed
        and there is a Block beneath it.
      - >-
        Blog The science behind how salt works lies in something known as the
        freezing point. Salt has a lower freezing point than water, which is why
        salt water does not freeze as quickly as pure water. This lower freezing
        point is what enables salt to melt ice and snow. When you spread salt on
        ice, it acts almost like a molecular sponge. It draws some of the frozen
        water molecules to it and, because of its lower freezing point, breaks
        the molecular bonds between them. These broken molecular bonds cause the
        ice to melt, creating a brine that encourages further melting.
      - >-
        - Design. Measuring in at 4.7 and 5.5-inches, the iPhone 6 and the
        iPhone 6 Plus dwarf previous iPhone models, but at the same time, the
        two phones are Apple's thinnest yet, measuring in at 6.9 and 7.1mm,
        respectively. For comparison's sake, the iPhone 5s was 7.6mm thick.hat's
        Next. The iPhone 6 and the iPhone 6 Plus were followed by the iPhone 6s
        and the iPhone 6s Plus, which launched on September 25, 2015. More
        information on Apple's latest iPhones can be found in our dedicated
        iPhone 6s roundup.
      - >-
        What happens if someone stops dialysis? For many people with kidney
        failure, dialysis or a kidney transplant enables them to live longer and
        enjoy their quality of life. However, this may not be the case for
        everyone and each person has the right to choose how — or even if — they
        want to receive treatment for chronic kidney disease.hat to expect once
        dialysis is stopped. Without dialysis, toxins build up in the blood,
        causing a condition called uremia. The patient will receive whatever
        medicines are necessary to manage symptoms of uremia and other medical
        conditions.
      - >-
        What happens to rocks when they are heated? Rating Newest Oldest. Best
        Answer: You heat it, it gets bigger. You heat it too fast, it will crack
        and break. You heat it slowly within the earth's crust, and subject it
        to pressure, the crystals within it will relign and it will
        metamorphose. You heat it strongly enough, and it will melt and becomne
        magma.
  - source_sentence: what is an archipelago?
    sentences:
      - >-
        - Save. Hardening steel with motor oil is a way of performing what is
        called the case hardening of steel. Pure steel is actually too soft for
        many applications. In order to put a hard layer on the steel, carbon
        must be fused at the molecular level into the top centimeter or so of
        the steel.
      - >-
        - An archipelago is a group, chain or cluster of islands surrounded by a
        large body of water, usually the open sea. + 85 others found this
        useful. Ben Ward. Edit. Share to: Bayley Dole. 1,609 Contributions. Avid
        reader of American History books, previous AP Government and AP US
        History student. Getting to Know Micronesia Micronesia is an archipelago
        that consists of 607 islands. 2  First colonized by Spain in the 17th
        century, the archipelago was ruled by multiple countries before ending
        up unde….
      - >-
        Island An island is an area of land smaller than a continent and
        entirely surrounded by water. Islands range from single rocks to huge
        landmasses, such as the island of Greenland. There are two main types of
        island—continental islands and oceanic islands. Islands are also found
        in rivers and lakes.
      - >-
        Malta (island) Malta is the largest of the three major islands that
        constitute the Maltese archipelago. It is sometimes referred to as
        Valletta for statistical purposes to distinguish the main island from
        the entire country. Malta is in the middle of the Mediterranean Sea
        directly south of Italy and north of Libya.The island is 27 kilometres
        (17 mi) long and 14.5 kilometres (9 mi) wide, with a total area of 246
        square kilometres (95 sq mi).The capital is Valletta, largest city is
        Qormi and the largest locality is Birkirkara.alta is the largest of the
        three major islands that constitute the Maltese archipelago. It is
        sometimes referred to as Valletta for statistical purposes to
        distinguish the main island from the entire country. Malta is in the
        middle of the Mediterranean Sea directly south of Italy and north of
        Libya.
      - >-
        Overview The Seychelles is synonymous with the world’s most beautiful
        islands! This archipelago is made up of over a hundred islands, each one
        different from the next, in the perfect azure of the Indian Ocean.
      - >-
        About the Galapagos Islands The Galapagos Archipelago is a cluster of
        some 13 volcanic islands and associated islets and rocks located just
        under the equator, about 600 miles (1000km) west of Ecuador in South
        America.
      - >-
        Heroin Heroin is a depressant drug, which means it slows down the
        messages travelling between your brain and body. Heroin belongs to a
        group of drugs known as ‘opioids’ that are from the opium poppy. 1.
      - >-
        - The Maldives, an Islamic Republic, is an archipelago of 1,191 islands
        in the Indian Ocean. Its culture is a blend of influences from southern
        India, Sri Lanka, East Africa and Arab countries.
  - source_sentence: enteroaggregative escherichia coli definition
    sentences:
      - >-
        5 Things You Need to Know About a Sudden Drop in Blood Pressure Rapid
        Changes in Health Can Cause an Acute Drop in Blood Pressure. Certain
        medical conditions can cause an acute drop in blood pressure. Blood loss
        from either an external injury or internal bleeding, for instance, will
        cause a drop in blood pressure. Excessive blood loss during menstruation
        could also be a cause.
      - >-
        Footloose (RAH) Footloose had it all. Highly intelligent valedictorian
        in high school, track team captain, gone to university on a state
        scholarship, an upbeat life anyone can be envious of. For some
        inexplicable reason, the man just went off the public radar, leaving his
        past and achievements.
      - >-
        E. coli E. coli. E. coli (Escherichia coli), species of bacterium that
        normally inhabits the stomach and intestines. When E. coli is consumed
        in contaminated water, milk, or food or is transmitted through the bite
        of a fly or other insect, it can cause gastrointestinal illness.
      - >-
        Escherichia coli Escherichia coli (IPA: [ˌɛ.ʃəˈɹɪ.kjə ˈkʰoʊ.laɪ]), is
        one of the well known and significant species of bacteria living as gut
        fauna in the lower intestines of mammals. The number of individual E.
        coli bacteria in the feces that a human excretes in one day averages
        between 100 billion and 10 trillion.
      - >-
        Enterotoxigenic E. coli  (ETEC) Enterotoxigenic E. coli (ETEC)
        Enterotoxigenic Escherichia coli (E. coli), or ETEC, is an important
        cause of bacterial diarrheal illness. Infection with ETEC is the leading
        cause of travelers' diarrhea and a major cause of diarrheal disease in
        lower-income countries, especially among children.
      - >-
        Escherichia coli noun, Bacteriology. 1. a species of rod-shaped,
        facultatively anaerobic bacteria in the large intestine of humans and
        other animals, sometimes pathogenic. Origin of Escherichia coli Expand.
        < New Latin, named after T. Escherich (died 1911), German physician; see
        -ia.
      - >-
        Preface Enteroaggregative Escherichia coli (EAEC), an increasingly
        recognized cause of diarrhea in children in developing countries, has
        been particularly associated with persistent diarrhea (more than 14
        days), a major cause of illness and death.
      - >-
        E.coli (Escherichia coli) Escherichia coli (abbreviated as E. coli) are
        bacteria found in the environment, foods, and intestines of people and
        animals.E. coli are a large and diverse group of bacteria. Although most
        strains of E. coli are harmless, others can make you sick.Some kinds of
        E. coli can cause diarrhea, while others cause urinary tract infections,
        respiratory illness and pneumonia, and other illnesses. More ».Learn
        about multistate outbreaks of E. coli.scherichia coli (abbreviated as E.
        coli) are bacteria found in the environment, foods, and intestines of
        people and animals.
  - source_sentence: who is the president united states today
    sentences:
      - >-
        Modes of Transmission Bloodborne pathogens such as HBV and HIV can be
        transmitted through contact with infected human blood and other
        potentially infectious body fluids such as: Semen Vaginal secretions
        Cerebrospinal fluid Synovial fluid Pleural fluid Peritoneal fluid
        Amniotic fluid Saliva (in dental procedures), and Any body fluid that is
        visibly contaminated with blood. It is important to know the ways
        exposure and transmission are most likely to occur in your particular
        situation, be it providing first aid to a student in the classroom,
        handling blood samples in the laboratory, or cleaning up blood from a
        hallway.
      - >-
        Barack Obama Barack Obama, Democrat, (born August 4, 1961 in Honolulu, 
        Hawaii) officially became the forty-fourth President of the United 
        States, on January 20, 2009 and was re-electe … d for a second term on 
        November 6, 2012. Barack Obama became president of the United States of
        America in  2009 and is the current president. Before he became
        president there  was George W. Bush and Bill Clinton.
      - >-
        Presidents Barack Hussein Obama II (US /bəˈrɑːk huːˈseɪn oʊˈbɑːmə/; born
        August 4, 1961) is an American politician who is the 44th and current
        President of the United States. He is the first African American to hold
        the office and the first president born outside the continental United
        States. Born in Honolulu, Hawaii, Obama is a graduate of Columbia
        University and Harvard Law School, where he was president of the Harvard
        Law Review. He was a community organizer in Chicago before earning his
        law degree.
      - >-
        - In The News Today. 1  Donald Trump hosted Sarah Palin, Kid Rock, and
        Ted Nugent at the White House. 2  Marine Le Pen has embraced being a
        woman, and that might just help her become the next big far-right world
        leader 3  🇫🇷. FYI: If you've agreed to Uber's terms of service you've
        signed away the right to sue the company 😬 🚕.
      - >-
        - It’s Labor Day, 2020. President Donald Trump is nearing the end of his
        first term in the White House — and getting ready for the final sprint
        in his re-election campaign.
      - >-
        Carbon dioxide Carbon dioxide is produced during the processes of decay
        of organic materials and the fermentation of sugars in bread, beer and
        winemaking. It is produced by combustion of wood and other organic
        materials and fossil fuels such as coal, peat, petroleum and natural
        gas.
      - >-
        - 1 Pope Francis will address the United Nations' biggest gathering of
        world leaders today. 2  Yesterday, he delivered a historic speech to the
        U.S. Congress. ›. 3  With a possible government shutdown looming, the
        U.S. Senate blocked an effort by Republicans to deny federal funding for
        Planned Parenthood. Pope Francis will address the United Nations'
        biggest gathering of world leaders today. 2  Yesterday, he delivered a
        historic speech to the U.S. Congress. ›. 3  With a possible government
        shutdown looming, the U.S. Senate blocked an effort by Republicans to
        deny federal funding for Planned Parenthood.
      - >-
        - Abraham Lincoln (12 February 1809 - 15 April 1865) was the 16th
        President of the United States.
datasets:
  - mixedbread-ai/rlhn-680k-msmarco-7negs-scored
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
co2_eq_emissions:
  emissions: 41.76368695617693
  energy_consumed: 0.156054759908292
  source: codecarbon
  training_type: fine-tuning
  on_cloud: false
  cpu_model: 13th Gen Intel(R) Core(TM) i7-13700K
  ram_total_size: 31.777088165283203
  hours_used: 0.794
  hardware_used: 1 x NVIDIA GeForce RTX 3090
model-index:
  - name: jhu-clsp/ettin-encoder-17m trained on RLHN MS MARCO using cmnrl
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: NanoMSMARCO
          type: NanoMSMARCO
        metrics:
          - type: cosine_accuracy@1
            value: 0.22
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.4
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.52
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.64
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.22
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.13333333333333333
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.10400000000000001
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.064
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.22
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.4
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.52
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.64
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.41264403293661234
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.34149206349206346
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.3591274789371428
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: NanoNFCorpus
          type: NanoNFCorpus
        metrics:
          - type: cosine_accuracy@1
            value: 0.3
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.36
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.46
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.5
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.3
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.22666666666666666
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.2
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.156
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.007796302222821822
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.01980804838125997
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.03461116468772874
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.045785191165931556
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.18317312912843284
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.35538888888888887
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.051374407713498646
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: NanoNQ
          type: NanoNQ
        metrics:
          - type: cosine_accuracy@1
            value: 0.24
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.34
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.36
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.46
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.24
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.11333333333333333
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.07200000000000001
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.05000000000000001
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.22
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.31
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.33
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.44
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.31950539918792653
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.29594444444444445
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.29368337934177885
            name: Cosine Map@100
      - task:
          type: nano-beir
          name: Nano BEIR
        dataset:
          name: NanoBEIR mean
          type: NanoBEIR_mean
        metrics:
          - type: cosine_accuracy@1
            value: 0.25333333333333335
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.3666666666666667
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.4466666666666666
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.5333333333333333
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.25333333333333335
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.15777777777777777
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.12533333333333335
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09000000000000001
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.14926543407427395
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.24326934946042
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.29487038822924294
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.37526173038864385
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.30510752041765726
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.3309417989417989
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.23472842199747346
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: rlhn msmarco eval
          type: rlhn-msmarco-eval
        metrics:
          - type: cosine_accuracy@1
            value: 0.36466666666666664
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.531
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.6013333333333334
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.686
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.36466666666666664
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.177
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.12026666666666665
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.0686
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.36466666666666664
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.531
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.6013333333333334
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.686
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.5178156881433742
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.4647850529100526
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.4730633839153687
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: rlhn msmarco test
          type: rlhn-msmarco-test
        metrics:
          - type: cosine_accuracy@1
            value: 0.2928
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.4698
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.5568
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.6617
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.2928
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.1566
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.11135999999999999
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.06616999999999998
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.2928
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.4698
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.5568
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.6617
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.46596220097347807
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.4046276587301606
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.4133624239973121
            name: Cosine Map@100

jhu-clsp/ettin-encoder-17m trained on RLHN MS MARCO using cmnrl

This is a sentence-transformers model finetuned from jhu-clsp/ettin-encoder-17m on the rlhn-680k-msmarco-7negs-scored dataset. It maps sentences & paragraphs to a 256-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: jhu-clsp/ettin-encoder-17m
Maximum Sequence Length: 7999 tokens
Output Dimensionality: 256 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- rlhn-680k-msmarco-7negs-scored
Language: en
License: apache-2.0

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': 7999, 'do_lower_case': False, 'architecture': 'ModernBertModel'})
  (1): Pooling({'word_embedding_dimension': 256, '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("tomaarsen/ettin-encoder-17m-cmnrl-lr2e-05-bs128")
# Run inference
queries = [
    "who is the president united states today",
]
documents = [
    '- It’s Labor Day, 2020. President Donald Trump is nearing the end of his first term in the White House — and getting ready for the final sprint in his re-election campaign.',
    '- Abraham Lincoln (12 February 1809 - 15 April 1865) was the 16th President of the United States.',
    'Barack Obama Barack Obama, Democrat, (born August 4, 1961 in Honolulu,  Hawaii) officially became the forty-fourth President of the United  States, on January 20, 2009 and was re-electe … d for a second term on  November 6, 2012. Barack Obama became president of the United States of America in  2009 and is the current president. Before he became president there  was George W. Bush and Bill Clinton.',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 256] [3, 256]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.4806, 0.6329, 0.7192]])

Evaluation

Metrics

Information Retrieval

Datasets: NanoMSMARCO, NanoNFCorpus and NanoNQ

Evaluated with InformationRetrievalEvaluator with these parameters:

{
    "query_prompt": "query: ",
    "corpus_prompt": "document: "
}

Metric	NanoMSMARCO	NanoNFCorpus	NanoNQ
cosine_accuracy@1	0.22	0.3	0.24
cosine_accuracy@3	0.4	0.36	0.34
cosine_accuracy@5	0.52	0.46	0.36
cosine_accuracy@10	0.64	0.5	0.46
cosine_precision@1	0.22	0.3	0.24
cosine_precision@3	0.1333	0.2267	0.1133
cosine_precision@5	0.104	0.2	0.072
cosine_precision@10	0.064	0.156	0.05
cosine_recall@1	0.22	0.0078	0.22
cosine_recall@3	0.4	0.0198	0.31
cosine_recall@5	0.52	0.0346	0.33
cosine_recall@10	0.64	0.0458	0.44
cosine_ndcg@10	0.4126	0.1832	0.3195
cosine_mrr@10	0.3415	0.3554	0.2959
cosine_map@100	0.3591	0.0514	0.2937

Nano BEIR

Dataset: NanoBEIR_mean

Evaluated with NanoBEIREvaluator with these parameters:

{
    "dataset_names": [
        "msmarco",
        "nfcorpus",
        "nq"
    ],
    "query_prompts": {
        "msmarco": "query: ",
        "nfcorpus": "query: ",
        "nq": "query: "
    },
    "corpus_prompts": {
        "msmarco": "document: ",
        "nfcorpus": "document: ",
        "nq": "document: "
    }
}

Metric	Value
cosine_accuracy@1	0.2533
cosine_accuracy@3	0.3667
cosine_accuracy@5	0.4467
cosine_accuracy@10	0.5333
cosine_precision@1	0.2533
cosine_precision@3	0.1578
cosine_precision@5	0.1253
cosine_precision@10	0.09
cosine_recall@1	0.1493
cosine_recall@3	0.2433
cosine_recall@5	0.2949
cosine_recall@10	0.3753
cosine_ndcg@10	0.3051
cosine_mrr@10	0.3309
cosine_map@100	0.2347

Information Retrieval

Datasets: rlhn-msmarco-eval and rlhn-msmarco-test

Evaluated with InformationRetrievalEvaluator with these parameters:

{
    "query_prompt_name": "query",
    "corpus_prompt_name": "document"
}

Metric	rlhn-msmarco-eval	rlhn-msmarco-test
cosine_accuracy@1	0.3647	0.2928
cosine_accuracy@3	0.531	0.4698
cosine_accuracy@5	0.6013	0.5568
cosine_accuracy@10	0.686	0.6617
cosine_precision@1	0.3647	0.2928
cosine_precision@3	0.177	0.1566
cosine_precision@5	0.1203	0.1114
cosine_precision@10	0.0686	0.0662
cosine_recall@1	0.3647	0.2928
cosine_recall@3	0.531	0.4698
cosine_recall@5	0.6013	0.5568
cosine_recall@10	0.686	0.6617
cosine_ndcg@10	0.5178	0.466
cosine_mrr@10	0.4648	0.4046
cosine_map@100	0.4731	0.4134

Training Details

Training Dataset

rlhn-680k-msmarco-7negs-scored

Dataset: rlhn-680k-msmarco-7negs-scored at 88f1f50
Size: 1,332,110 training samples
Columns: query, positive, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, and scores

Approximate statistics based on the first 1000 samples:

	query	positive	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	scores
type	string	string	string	string	string	string	string	string	string	list
details	min: 4 tokens mean: 9.12 tokens max: 24 tokens	min: 17 tokens mean: 90.52 tokens max: 274 tokens	min: 18 tokens mean: 85.33 tokens max: 208 tokens	min: 20 tokens mean: 83.66 tokens max: 226 tokens	min: 21 tokens mean: 84.67 tokens max: 258 tokens	min: 17 tokens mean: 82.88 tokens max: 208 tokens	min: 21 tokens mean: 83.68 tokens max: 288 tokens	min: 19 tokens mean: 82.63 tokens max: 207 tokens	min: 19 tokens mean: 83.97 tokens max: 223 tokens	size: 8 elements

Samples:

query	positive	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	scores
`is formaldehyde water soluble`	`Why is formaldehyde a good preservative? Formaldehyde is a cheap preservative that mixes easily with water but will not mix with oil or grease, so it is fairly common to find it in watery concoctions like shampoo, conditioner, shower gel, liquid hand wash and even children's bubble bath.`	`Facts About Formaldehyde Formaldehyde is found in: 1 Resins used in the manufacture of composite wood products (i.e., hardwood plywood, particleboard and medium-density fiberboard) 2 Building materials and insulation. 3 Household products such as glues, permanent press fabrics, paints and coatings, lacquers and finishes, and paper products.`	`Tosylamide/Formaldehyde Resin Tosylamide/Formaldehyde Resin is used in nail enamel products. It is a soft resin material which, in conjunction with nitrocellulose, forms tough, shiny and durable films on nails.osylamide/Formaldehyde Resin is a synthetic resin, sometimes called Toluenesulfonamide/Formaldehyde Resin. It modifies the film-forming characteristics of Nitrocellulose type polymers.`	`- Formaldehyde is a colorless gas with a rather strong odor, made up of a carbon atom, two hydrogen atoms and an atom of oxygen, CH2O. It has very potent antibacterial and antifungal properties. Many textile industries use formaldehyde in their production processes.`	`Formaldehyde Formaldehyde (systematic name methanal), is a naturally occurring organic compound with the formula CH 2 O (H-CHO). It is the simplest of the aldehydes (R-CHO). The common name of this substance comes from its similarity and relation to formic acid. Formaldehyde is an important precursor to many other materials and chemical compounds.`	`Dissolving tetracycline in water? Well though I have no experience with your compound, however, whenever we had to dissolve non-polar in water we dissolved the compound in a very little amount of compatible solvent (eg acetone, alcohol etc) thereafter, it become miscible in water. ... ·. Hello! Thanks everyone for the answers.`	`PowerEdge VRTX CMC Licensing Details PowerEdge VRTX CMC Licensing Details. As I noted in my original posting of the PowerEdge VRTX, the Chassis Management Controller, or CMC, is available with an “Express” license and an “Enterprise” license. To help you determine which would be best suited for your environment, check out the chart below.`	`- About Poe's Short Stories. These stories represent the highest achievements in the literary genre of the gothic horror story. By gothic, one means that the author emphasizes the grotesque, the mysterious, the desolate, the horrible, the ghostly, and, ultimately, the abject fear that can be aroused in either the reader or in the viewer.`	`[0.7772998508568416, 0.0010420335130684381, 0.004322105866822573, 0.06008664344151498, 0.07159119944455247, ...]`
`what presidents replaced assassinated presidents`	`Which four presidents of the United States were assassinated? There have been eighteen attempts to kill sitting and former United States Presidents and Presidents-Elect. Four attempts on sitting Presidents have succeeded; the 16th, 20th, 25th and 35th US Presidents were all assassinated in office, and two others were injured. Successful assassinations: 1 -- Abraham Lincoln.`	`- Register Now. 1 Solution Finder Answers to your top questions about your Ford. 2 Post a comment Find support among your fellow driver. 3 Live chat Get one-on-one support right now. Call for support Speak to a Ford 1 representative. Find a dealer Find a Ford dealership in your area.`	`Who Knew? U.S. Presidential Trivia Assassination and Death. Four sitting Presidents have been assassinated while in office: Abraham Lincoln, James Abram Garfield (1881), William McKinley (1897-1901), and John F. Kennedy (1961-63).`	`- Prince Y Johnson is better known as the man who presided over the torture and assassination of the then Liberian President, Samuel Doe, in September of 1990. Even by the standards of the time it was a particularly brutal slaying. Before he was killed, Doe was mutilated, his ears sliced off.`	`Who was the first vice president to take presidency after a death? John Tyler was the first Vice-President to take over after the death of a sitting President. William Henry Harrison was sworn in as President on March 4, 1841.`	`How Is a New Vice President Chosen? Advertisement. The 25th Amendment, ratified in 1967, solved the problem of how to replace a vice president. If Cheney resigned, President Bush would select a nominee for the vice presidency. Then the nominee would have to be approved by majority votes in both the House and the Senate.`	`Campylobacteriosis - Causes and Symptoms Some persons who are infected with Campylobacter don't have any symptoms at all. In persons with compromised immune systems, Campylobacter occasionally spreads to the bloodstream and causes a serious life-threatening infection.`	`- Dotplot Overview. As you might guess, a dotplot is made up of dots plotted on a graph. Here is how to interpret a dotplot. 1 Each dot represents a specific number of observations from a set of data. ( 2 Unless otherwise indicated, assume that each dot represents one observation.`	`[0.34864513533394575, 0.000999042306106755, 0.679178699175393, 0.003032753157394305, 0.01495708659314999, ...]`
`what military base is the largest in the us`	`What are the 10 largest U.S. military bases? , Proficient in the use of Google. The top 5 largest military bases in the world belong to the US. sorry for only being able to find the information on 5 bases. 1. Fort Bragg (population 238,646 & 163,000 acres. 2. Fort Campbell (population 234,914 & 105,000 acres) 3. Fort Hood (population 217,003 & 214,968 acres)`	What Are Some of the Largest Marine Corps Bases? Southern California's Camp Pendleton, which spans more than 125,000 acres, is the Marine Corps' largest expeditionary training facility on the West Coast. Camp Lejeune, situated on approximately 156,000 acres including about 11 miles of beach, is the Marine Corps' largest East Coast base. The base houses several training schools and is home for about 47,000 sailors and Marines who belong to the II Marine Expeditionary Force, an air and ground task force.	`The List: The Six Most Important U.S. Military Bases The best part: unlike other large bases in the region, Guam is U.S. territory. Balad Air Base/Camp Anaconda, Iraq. The base: Most prominent of the enduring bases being constructed in Iraq, Balad is located just north of Baghdad. It is one of the busiest airfields in the country, accommodating both Air Force fighters as well as transport aircraft.`	`What Are Some of the Largest Marine Corps Bases? A: America's largest Marine Corps bases include North Carolina's Marine Corps Base Camp Lejeune, and California's Marine Corps Air Ground Combat Center Twentynine Palms and Marine Corps Base Camp Pendleton. Commissioned during the Korean War in 1952, Twentynine Palms, the largest Marine Corps base in the world, exceeds 998 square miles.`	`Seeing shared threats, Turkey sets up military base in Qatar One of the world's wealthiest countries, Qatar is home to the largest U.S. air base in the Middle East, Al Udeid, where around 10,000 military personnel are stationed. (Reporting by Tom Finn; Editing by Mark Trevelyan)`	`Fort Hood: The largest US military base in the world Fort Hood is the largest active-duty, armoured US military base in the world and at the time of the shooting housed around 65,000 people. Sgt. Fanuaee Vea embraces Pvt. Savannah Green while trying to reach friends and family outside Fort Hood, Texas Photo: GETTY. 7:00AM GMT 06 Nov 2009.`	- Age. Standard Age-Most Soldiers who have been issued a NOE will become eligible for retired pay upon their 60th birthday. Service beyond age 60 - Soldiers remaining in an active status with the ARNG beyond the age of 60 will become eligible upon receipt of NOE and transfer to the retired reserve.on-regular retirement is the term used for ARNG and USAR Soldiers that have completed 20 or more years of qualifying service for retired pay upon reaching age 60 or, in some cases, sooner depending on their computed eligibility date.	`Can I Use Bluetooth Headphones with a Bluetooth A/V Receiver? They Don't Pair! RF-based wireless headphones typically have a pair of RCA jacks as input. Sometimes they'll have a standard 3.5mm (1/8) headphone jack instead. In either case, you can connect the receiver's headphone out jack to the wireless headphone transmitter and that will transmit sound to the wireless headphones.`	`[0.880797071719777, 0.754914975839675, 0.4378235137849365, 0.8519527944504226, 0.5312093733737563, ...]`

Loss: CachedMultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "mini_batch_size": 16,
    "gather_across_devices": false
}

Evaluation Dataset

rlhn-680k-msmarco-7negs-scored

Dataset: rlhn-680k-msmarco-7negs-scored at 88f1f50
Size: 3,000 evaluation samples
Columns: query, positive, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, and scores

Approximate statistics based on the first 1000 samples:

	query	positive	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	scores
type	string	string	string	string	string	string	string	string	string	list
details	min: 4 tokens mean: 9.17 tokens max: 36 tokens	min: 26 tokens mean: 88.94 tokens max: 216 tokens	min: 23 tokens mean: 86.2 tokens max: 213 tokens	min: 23 tokens mean: 86.38 tokens max: 237 tokens	min: 20 tokens mean: 85.61 tokens max: 228 tokens	min: 19 tokens mean: 83.72 tokens max: 234 tokens	min: 17 tokens mean: 85.51 tokens max: 510 tokens	min: 19 tokens mean: 83.5 tokens max: 197 tokens	min: 21 tokens mean: 85.41 tokens max: 207 tokens	size: 8 elements

Samples:

query	positive	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	scores
`benefit payments pensions`	`Benefit Payments and Direct Deposit Benefit Payments and Direct Deposit. When your pension begins, you receive a benefit payment each month (unless you receive a mandatory cash-out payment or you have elected a voluntary cash-out of your benefit). You can elect payment by mail or through electronic direct deposit. We encourage benefit recipients to sign up for direct deposit.`	`Defined benefit pension plan Defined benefit pension plan. A defined benefit pension plan is a type of pension plan in which an employer/sponsor promises a specified monthly benefit on retirement that is predetermined by a formula based on the employee's earnings history, tenure of service and age, rather than depending directly on individual investment returns.`	Your Pension and Social Security Benefits Pension plans are offered and maintained by both private- and public-sector employers. You, and many other retirees, may collect both Social Security benefits and a monthly pension check. Receiving a pension doesn't automatically reduce your Social Security benefits. The key factor in determining whether your Social Security benefits will be affected by your pension is whether you receive a pension from government or other non-covered work. Receiving a Private Pension	`State Pension (United Kingdom) The State Pension is a contribution-based benefit, and depends on an individual's National Insurance (NI) contribution history. For someone with the 30 qualifying years (years in which NI contributions were paid), it is payable at a flat rate of £119.30 a week (up to April 2016). 35 years contribution is needed to get the full new state pension from 6 April 2016. A smaller, pro-rata, pension is paid to someone with fewer qualifying years.`	`Canada Pension Plan Survivor Benefits Survivor’s Pension. The survivor’s pension is a monthly amount paid to the deceased’s spouse or common-law partner. This amount is entered on line 114. For children or spouses to be eligible to receive survivor’s benefits, the deceased contributor must have made contributions for a minimum of three years.`	Inflation link on final salary pensions faces axe Rip-off pensions: Charges 'to be capped at 0.75pc'. The proposal for final salary schemes, dubbed flexible defined benefits aims to transfer more risk to savers by removing the requirement for schemes to increase pension payments in line with inflation, known as indexation.Employers would still provide a set level of income in retirement however.he proposal for final salary schemes, dubbed flexible defined benefits aims to transfer more risk to savers by removing the requirement for schemes to increase pension payments in line with inflation, known as indexation. Employers would still provide a set level of income in retirement however.	- Automated Clearing House (ACH) is an electronic network for financial transactions in the United States. ACH processes large volumes of credit and debit transactions in batches.ACH credit transfers include direct deposit, payroll and vendor payments.CH is a computer-based clearing and settlement facility established to process the exchange of electronic transactions between participating depository institutions. Rules and regulations that govern the ACH network are established by NACHA and the Federal Reserve.	`- The saturated vapor of inorganic compounds that evaporate in the temperature range 700–3000 K generally contains species formed by the dissociation of the base compound and a number of associates. Many interesting and unexpected examples can be found in the extensive database on the species in inorganic vapors.`	`[0.9706877722165251, 0.754914975839675, 0.5312093585306514, 0.6513548511303336, 0.0953494648991095, ...]`
`how big a macaw can get`	`Macaw Care The largest Macaw, and also the largest of all parrots, is the impressive Hyacinth Macaw at 36 in length with a wingspan of up to 42, and weighing up to 1550 grams. The smallest of the macaws is the Hahn's Macaw, reaching not quite 12 (30 cm) in length. Types of Macaws.`	`How big can a bobcat get? I have seen bobcats as much as 47.3 pounds where I live, which is in Virginia. A cat this big will exceed 4 feet in length and a foot or so at the shoulder. There is some vari…ation among the dozen subspecies. A link can be found below.`	`How long Do Parrots Live â Lifespan of Parrots Macaws are by far one of the longest living parrots. They have a lifespan of about 80 – 90 years. Some have even lived up to 100 years. Macaws are found in South America, Mexico, and Caribbean. Now you can well imagine how long do parrots live!`	`Scarlet macaw Scarlet macaw. The scarlet macaw (Ara macao) is a large red, yellow, and blue South American parrot, a member of a large group of Neotropical parrots called macaws. It is native to humid evergreen forests of tropical South America.`	`How Long Can a Large Parrot Live? Since they are the largest parrots, macaws also have the longest life expectancies. If a young macaw joins your family, it probably will outlive you, as healthy macaws can survive 60 to 80 years. Even the smaller macaws, like the Hahns macaw, can live up 60 years. If you want to spend your life with a colorful, feathered friend, a macaw might be a good choice.`	`HOW BIG CAN PARROTS GET? They grow fairly quickly (not. They grow fairly quickly (not nearly as quickly as most other big cichlids) for the first 2-3 years, but once they hit 3-5 their growth rate plummets. They usually top out under 8 but the parrots that lean more toward KK parrots can reach that size or even larger.`	`What is a badminton racquet made of? a badminton raquet is the peice of eqiupment that people (that are playing badmintom) use to hit the birdie/shuttlecock. - Chow. badminton racquet sizes is never over the size of 680mm in height and 230mm in width.`	`Big Rapids, Michigan Population: Census 2010 and 2000 Interactive Map, Demographics, Statistics, Quick Facts Population by Hispanic or Latino Origin (of any race), 2000 Census-Grand Rapids, Michigan.`	`[0.977022627413796, 0.004345778821719373, 0.1329642264548166, 0.08035747763179142, 0.7772998508568416, ...]`
`boundary waters canoe area wilderness fur post`	`Boundary Waters Canoe Area Wilderness (Reservations), MN The Boundary Waters Canoe Area Wilderness (BWCAW) is located in the northern third of the Superior National Forest in northeastern Minnesota. Approximately 1.1 million acres in size, it extends nearly 150 miles along the international boundary adjacent to Canada's Quetico Provincial Park and is bordered on the west by Voyageurs National Park.`	ZipRiderÂ® Icy Strait Point, Alaska The typical ride takes 1 ½ minutes and concludes on the beach at Icy Strait. Icy Strait Point is a cruise line port of call, located on the Chichagof Island, 1.5 miles away from the town of Hoonah, Alaska. Chichagof Island is located 50 miles west of Juneau. The Icy Strait Point ZipRider® opened in May of 2007 and has been the anchor attraction for Icy Strait Point ever since. People come from all over the country and world to experience the world renowned Icy Strait Point ZipRider® adventure and adrenaline rush!	`Fort Wilderness Resort and Campground Fact Sheet In fact, Fort Wilderness is the only pet friendly resort on Disney World property. Request a pet site when making reservations. Pets are allowed in Premium, Preferred and Full Hook-Up campsites in loops 1600 through 1900, 300, 700 and 800, but must be kept in those areas.`	`- 357 Redhawk. This is a discussion on 357 Redhawk within the Ruger Double Action forums, part of the Pistol & Revolver Forum category; Hello all. New here and have a few questions. I have searched the forums and learned alot, but still have a few.`	`Cirque Lake - Banff National Park - Hiking Alberta Rockfall to the shore of Cirque Lake above Waterfowl Lakes Campground along Icefields Parkway in Banff National Park, Alberta, Canada. Unlimited opportunities for artistic photographs at Cirque Lake above Waterfowl Lakes Campground along Icefields Parkway in Banff National Park, Alberta, Canada.`	`- Friends of the Boundary Waters Wilderness. Urgent action is needed to curb the effects of human activity on our planet's climate and prepare for impacts of #climatechange. Ignoring climate change will endanger millions of lives and livelihoods across the planet as well as threaten national security, the economy, and of course the environment.`	Commonwealth of Independent States The Commonwealth of Independent States (CIS) is the international organization, or alliance of countries that used to form the Soviet Union. On December 8 1991, leaders of Russia, Belarus, and Ukraine met at the nature reserve called Belovezhskaya Pushcha in Belarus. They discussed the creation of the Commonwealth of Independent States (CIS). They reached an agreement about many points and signed a document creating the CIS. One of these points stated that the alliance would be open to all the republics of the Soviet Union.	`- It could cost less than $20 for something that can be buffed out. $40 for a mediocre touch up job (or $10 for one you can do yourself), up to $500 to repaint and blend a complete bumper, even more on a luxury car where exact paint matches may be hard to find and expensive.`	`[0.9626731094431571, 0.001340751940071589, 0.0029596639358424365, 0.0013305401297137621, 0.0018691848506214679, ...]`

Loss: CachedMultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "mini_batch_size": 16,
    "gather_across_devices": false
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 128
per_device_eval_batch_size: 128
learning_rate: 2e-05
num_train_epochs: 0.05
warmup_ratio: 0.1
bf16: True
prompts: {'query': 'query', 'positive': 'document', 'negative_1': 'document', 'negative_2': 'document', 'negative_3': 'document', 'negative_4': 'document', 'negative_5': 'document', 'negative_6': 'document', 'negative_7': 'document'}
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: 128
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
torch_empty_cache_steps: None
learning_rate: 2e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 0.05
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.1
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
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: True
fp16: False
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: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
hub_revision: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
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
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
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: {'query': 'query', 'positive': 'document', 'negative_1': 'document', 'negative_2': 'document', 'negative_3': 'document', 'negative_4': 'document', 'negative_5': 'document', 'negative_6': 'document', 'negative_7': 'document'}
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss	NanoMSMARCO_cosine_ndcg@10	NanoNFCorpus_cosine_ndcg@10	NanoNQ_cosine_ndcg@10	NanoBEIR_mean_cosine_ndcg@10	rlhn-msmarco-eval_cosine_ndcg@10	rlhn-msmarco-test_cosine_ndcg@10
0.0096	100	5.1901	-	-	-	-	-	-	-
0.0101	105	-	2.4238	0.3045	0.1696	0.2506	0.2416	0.4009	-
0.0192	200	2.2564	-	-	-	-	-	-	-
0.0202	210	-	2.0731	0.3990	0.1838	0.2834	0.2887	0.4811	-
0.0288	300	2.0837	-	-	-	-	-	-	-
0.0303	315	-	1.9839	0.4068	0.1827	0.2954	0.2950	0.5052	-
0.0384	400	1.9895	-	-	-	-	-	-	-
0.0404	420	-	1.9464	0.4197	0.1827	0.3050	0.3025	0.5148	-
0.0480	500	1.9635	-	-	-	-	-	-	-
-1	-1	-	-	0.4126	0.1832	0.3195	0.3051	0.5178	0.4660

Environmental Impact

Carbon emissions were measured using CodeCarbon.

Energy Consumed: 0.156 kWh
Carbon Emitted: 0.042 kg of CO2
Hours Used: 0.794 hours

Training Hardware

On Cloud: No
GPU Model: 1 x NVIDIA GeForce RTX 3090
CPU Model: 13th Gen Intel(R) Core(TM) i7-13700K
RAM Size: 31.78 GB

Framework Versions

Python: 3.11.6
Sentence Transformers: 5.2.0.dev0
Transformers: 4.56.0.dev0
PyTorch: 2.7.1+cu126
Accelerate: 1.6.0
Datasets: 3.6.0
Tokenizers: 0.21.1

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",
}

CachedMultipleNegativesRankingLoss

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}