static-retrieval-mrl-en-v1

Static Embeddings with BERT uncased tokenizer finetuned on various datasets

This is a sentence-transformers model trained on the gooaq, msmarco, squad, s2orc, allnli, paq, trivia_qa, msmarco_10m, swim_ir, pubmedqa, miracl, mldr and mr_tydi datasets. It maps sentences & paragraphs to a 1024-dimensional dense vector space and is designed to be used for semantic search.

Read our Static Embeddings blogpost to learn more about this model and how it was trained.

• 0 Active Parameters: This model does not use any active parameters, instead consisting exclusively of averaging pre-computed token embeddings.
• 100x to 400x faster: On CPU, this model is 100x to 400x faster than common options like all-mpnet-base-v2. On GPU, it's 10x to 25x faster.
• Matryoshka: This model was trained with a Matryoshka loss, allowing you to truncate the embeddings for faster retrieval at minimal performance costs.
• Evaluations: See Evaluations for details on performance on NanoBEIR, embedding speed, and Matryoshka dimensionality truncation. In short, this model is 87.4% as performant as the commonly used all-mpnet-base-v2.
• Training Script: See train.py for the training script used to train this model from scratch.

See static-similarity-mrl-multilingual-v1 for a general-purpose multilingual static embedding model. It's been trained for semantic textual similarity, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer

<!-- - **Base model:** [Unknown](https://huggingface.co/unknown) -->

• Maximum Sequence Length: inf tokens
• Output Dimensionality: 1024 tokens
• Similarity Function: Cosine Similarity
• Training Datasets:
  • gooaq
  • msmarco
  • squad
  • s2orc
  • allnli
  • paq
  • trivia_qa
  • msmarco_10m
  • swim_ir
  • pubmedqa
  • miracl
  • mldr
  • mr_tydi
• 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): StaticEmbedding(
    (embedding): EmbeddingBag(30522, 1024, mode='mean')
  )
)

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/static-retrieval-mrl-en-v1")
# Run inference
sentences = [
    'Gadofosveset-enhanced MR angiography of carotid arteries: does steady-state imaging improve accuracy of first-pass imaging?',
    'To evaluate the diagnostic accuracy of gadofosveset-enhanced magnetic resonance (MR) angiography in the assessment of carotid artery stenosis, with digital subtraction angiography (DSA) as the reference standard, and to determine the value of reading first-pass, steady-state, and "combined" (first-pass plus steady-state) MR angiograms.',
    'In a longitudinal study we investigated in vivo alterations of CVO during neuroinflammation, applying Gadofluorine M- (Gf) enhanced magnetic resonance imaging (MRI) in experimental autoimmune encephalomyelitis, an animal model of multiple sclerosis. SJL/J mice were monitored by Gadopentate dimeglumine- (Gd-DTPA) and Gf-enhanced MRI after adoptive transfer of proteolipid-protein-specific T cells. Mean Gf intensity ratios were calculated individually for different CVO and correlated to the clinical disease course. Subsequently, the tissue distribution of fluorescence-labeled Gf as well as the extent of cellular inflammation was assessed in corresponding histological slices.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

This model was trained with Matryoshka loss, allowing this model to be used with lower dimensionalities with minimal performance loss (See Matryoshka Evaluations for evaluations). Notably, a lower dimensionality allows for much faster and cheaper information retrieval. You can specify a lower dimensionality with the truncate_dim argument when initializing the Sentence Transformer model:

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("tomaarsen/static-retrieval-mrl-en-v1", truncate_dim=256)
embeddings = model.encode([
    "what is the difference between chronological order and spatial order?",
    "can lavender grow indoors?"
])
print(embeddings.shape)
# => (2, 256)

<!-- ### Direct Usage (Transformers) <details><summary>Click to see the direct usage in Transformers</summary> </details> --> <!-- ### Downstream Usage (Sentence Transformers) You can finetune this model on your own dataset. <details><summary>Click to expand</summary> </details> --> <!-- ### Out-of-Scope Use *List how the model may foreseeably be misused and address what users ought not to do with the model.* -->

Evaluation

Metrics

Information Retrieval

• Datasets: NanoClimateFEVER, NanoDBPedia, NanoFEVER, NanoFiQA2018, NanoHotpotQA, NanoMSMARCO, NanoNFCorpus, NanoNQ, NanoQuoraRetrieval, NanoSCIDOCS, NanoArguAna, NanoSciFact and NanoTouche2020
• Evaluated with <code>InformationRetrievalEvaluator</code>

Nano BEIR

• Dataset: NanoBEIR_mean
• Evaluated with <code>NanoBEIREvaluator</code>

We've evaluated sentence-transformers/static-retrieval-mrl-en-v1 on NanoBEIR and plotted it against the inference speed computed on a RTX 3090 and i7-13700K. For the inference speed tests, we calculated the number of computed query embeddings of the GooAQ dataset per second, either on CPU or GPU.

We evaluate against 3 types of models:

1. Attention-based dense embedding models, e.g. traditional Sentence Transformer models like all-mpnet-base-v2, bge-base-en-v1.5, and gte-large-en-v1.5.

2. Static Embedding-based models, e.g. static-retrieval-mrl-en-v1, potion-base-8M, M2V_base_output, and glove.6B.300d.

3. Sparse bag-of-words model, BM25, often a strong baseline.

   <details><summary>Click to expand BM25 implementation details</summary>

   I relied on the highly efficient bm25s implementation, using model.get_scores() on tokens after tokenization and stemming with the English PyStemmer.

   </details>

NOTE: Many of the attention-based dense embedding models are finetuned on the training splits of the (Nano)BEIR evaluation datasets. This gives the models an unfair advantage in this benchmark and can result in lower downstream performance on real retrieval tasks.

static-retrieval-mrl-en-v1 is purposefully not trained on any of these datasets.

GPU

CPU

We can draw some notable conclusions from these figures:

1. static-retrieval-mrl-en-v1 outperforms all other Static Embedding models.
2. static-retrieval-mrl-en-v1 is the only Static Embedding model to outperform BM25.
3. static-retrieval-mrl-en-v1 is
   • 87.4% as performant as the commonly used all-mpnet-base-v2,
   • 24x faster on GPU,
   • 397x faster on CPU.
4. static-retrieval-mrl-en-v1 is quicker on CPU than on GPU: This model can run extraordinarily quickly everywhere, including consumer-grade PCs, tiny servers, phones, or in-browser.

Matryoshka Evaluations

We experimented with the results on NanoBEIR performance when we performed Matryoshka-style dimensionality reduction by truncating the output embeddings to a lower dimensionality.

These findings show that reducing the dimensionality by e.g. 2x only has a 1.47% reduction in performance (0.5031 NDCG@10 vs 0.4957 NDCG@10), while realistically resulting in a 2x speedup in retrieval speed.

<!-- ## Bias, Risks and Limitations *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.* --> <!-- ### Recommendations *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.* -->

Training Details

Training Datasets

<details><summary>gooaq</summary>

• Dataset: gooaq at b089f72
• Size: 3,012,496 training samples
• Columns: <code>question</code> and <code>answer</code>
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	<ul><li>min: 18 characters</li><li>mean: 43.23 characters</li><li>max: 96 characters</li></ul>	<ul><li>min: 55 characters</li><li>mean: 253.36 characters</li><li>max: 371 characters</li></ul>

• Samples:

  question	answer
  <code>what is the difference between broilers and layers?</code>	<code>An egg laying poultry is called egger or layer whereas broilers are reared for obtaining meat. So a layer should be able to produce more number of large sized eggs, without growing too much. On the other hand, a broiler should yield more meat and hence should be able to grow well.</code>
  <code>what is the difference between chronological order and spatial order?</code>	<code>As a writer, you should always remember that unlike chronological order and the other organizational methods for data, spatial order does not take into account the time. Spatial order is primarily focused on the location. All it does is take into account the location of objects and not the time.</code>
  <code>is kamagra same as viagra?</code>	<code>Kamagra is thought to contain the same active ingredient as Viagra, sildenafil citrate. In theory, it should work in much the same way as Viagra, taking about 45 minutes to take effect, and lasting for around 4-6 hours. However, this will vary from person to person.</code>

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>msmarco</summary>

• Dataset: msmarco at 84ed2d3
• Size: 502,939 training samples
• Columns: <code>query</code>, <code>positive</code>, and <code>negative</code>
• Approximate statistics based on the first 1000 samples:

  	query	positive	negative
  type	string	string	string
  details	<ul><li>min: 11 characters</li><li>mean: 33.26 characters</li><li>max: 197 characters</li></ul>	<ul><li>min: 96 characters</li><li>mean: 356.24 characters</li><li>max: 1006 characters</li></ul>	<ul><li>min: 68 characters</li><li>mean: 327.52 characters</li><li>max: 995 characters</li></ul>

• Samples:

  query	positive	negative
  <code>when was the sullivan acts</code>	<code>Sullivan Act Tim Sullivan, a major Irish criminal passed the Sullivan Act in 1911 to help his constituents rob strangers or to help them against Italian incomers. That is the crux of story that goes with a very early gun control law.</code>	<code>Sullivan Act Tim Sullivan, a major Irish criminal passed the Sullivan Act in 1911 to help his constituents rob strangers or to help them against Italian incomers. That is the crux of story that goes with a very early gun control law.</code>
  <code>can lavender grow indoors</code>	<code>Growing Lavender Indoors. People ALWAYS ask if you can grow lavender indoors. Well, you can, but most Lavender does best outside. Here is our winter experiment to show you what it would look like. This is one of our 4 Lavender Babies from Fall 2010. Our test specimen is L. x intermedia 'Grosso'.</code>	<code>Lavender can be grown indoors with a bit of effort to keep it in the conditions it loves to thrive. First off begin with choosing a variety that is better able to tolerate the conditions inside a home. To successfully grow Lavender indoors you need to create optimal growing conditions which is hard to do inside a house.</code>
  <code>what kind of barley do you malt</code>	<code>Barley is a wonderfully versatile cereal grain with a rich nutlike flavor and an appealing chewy, pasta-like consistency. Its appearance resembles wheat berries, although it is slightly lighter in color. Sprouted barley is naturally high in maltose, a sugar that serves as the basis for both malt syrup sweetener.</code>	<code>Specialty grains that can be used in this way are usually barley, malted or unmalted, that has been treated differently at the malting company. Crystal malt is one of the specialty grains. It is available in a whole range of colors, from 20 to 120 Lovibond. Crystal malt is malted barley that is heated while wet.</code>

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>squad</summary>

• Dataset: squad at d84c8c2
• Size: 87,599 training samples
• Columns: <code>question</code> and <code>answer</code>
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	<ul><li>min: 14 characters</li><li>mean: 59.66 characters</li><li>max: 150 characters</li></ul>	<ul><li>min: 156 characters</li><li>mean: 769.53 characters</li><li>max: 3706 characters</li></ul>

• Samples:

  question	answer
  <code>What did Business Insider call San Diego in 2013?</code>	<code>San Diego was ranked as the 20th-safest city in America in 2013 by Business Insider. According to Forbes magazine, San Diego was the ninth-safest city in the top 10 list of safest cities in the U.S. in 2010. Like most major cities, San Diego had a declining crime rate from 1990 to 2000. Crime in San Diego increased in the early 2000s. In 2004, San Diego had the sixth lowest crime rate of any U.S. city with over half a million residents. From 2002 to 2006, the crime rate overall dropped 0.8%, though not evenly by category. While violent crime decreased 12.4% during this period, property crime increased 1.1%. Total property crimes per 100,000 people were lower than the national average in 2008.</code>
  <code>What did the Spanish call this region?</code>	<code>The name Montana comes from the Spanish word Montaña, meaning "mountain", or more broadly, "mountainous country". Montaña del Norte was the name given by early Spanish explorers to the entire mountainous region of the west. The name Montana was added to a bill by the United States House Committee on Territories, which was chaired at the time by Rep. James Ashley of Ohio, for the territory that would become Idaho Territory. The name was successfully changed by Representatives Henry Wilson (Massachusetts) and Benjamin F. Harding (Oregon), who complained that Montana had "no meaning". When Ashley presented a bill to establish a temporary government in 1864 for a new territory to be carved out of Idaho, he again chose Montana Territory. This time Rep. Samuel Cox, also of Ohio, objected to the name. Cox complained that the name was a misnomer given that most of the territory was not mountainous and that a Native American name would be more appropriate than a Spanish one. Other names such as...</code>
  <code>Small missiles were designed that could be mounted on what?</code>	<code>As this process continued, the missile found itself being used for more and more of the roles formerly filled by guns. First to go were the large weapons, replaced by equally large missile systems of much higher performance. Smaller missiles soon followed, eventually becoming small enough to be mounted on armored cars and tank chassis. These started replacing, or at least supplanting, similar gun-based SPAAG systems in the 1960s, and by the 1990s had replaced almost all such systems in modern armies. Man-portable missiles, MANPADs as they are known today, were introduced in the 1960s and have supplanted or even replaced even the smallest guns in most advanced armies.</code>

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>s2orc</summary>

• Dataset: s2orc at 8cfc394
• Size: 90,000 training samples
• Columns: <code>title</code> and <code>abstract</code>
• Approximate statistics based on the first 1000 samples:

  	title	abstract
  type	string	string
  details	<ul><li>min: 31 characters</li><li>mean: 80.02 characters</li><li>max: 185 characters</li></ul>	<ul><li>min: 84 characters</li><li>mean: 635.31 characters</li><li>max: 1023 characters</li></ul>

• Samples:

  title	abstract
  <code>Modeling Method of Flow Diversion of the Three Outlets in Jingjiang Reach Under Unsteady Flow Conditions</code>	<code>The Yangtze River Flood Protection Physical Model is built under the financial support of World Bank loan.Based on theoretical analysis and experimental study,a modeling method of flow diversion of the three outlets in Jingjiang Reach under unsteady flow conditions was established for the model.Validation tests under both steady and unsteady flow conditions manifested that with this modeling method,the experimental flow diversion proves to be consistent with that of the prototype and therefore meets the requirements for precision.Being validated,this modeling method has been applied to Yangtze River Flood Protection Physical Model to study the flood routing features in Jingjiang reach.</code>
  <code>Enlightening on medical administration by clinical governance in British</code>	<code>Medical quality and safety were the responsibilities of medical system in view of British clinical governance. Medical regulation institutes were considered to be built and be authorized regulation rights. British medical administration was introduced and its enlightening in China was mentioned.</code>
  <code>APPLICATION OF A FUZZY MULTI-CRITERIA DECISION-MAKING MODEL FOR SHIPPING COMPANY PERFORMANCE EVALUATION</code>	<code>Combining fuzzy set theory, Analytic Hierarchy Process (AHP) and concept of entropy, a fuzzy Multiple Criteria Decision-Making (MCDM) model for shipping company performance evaluation is proposed. First, the AHP is used to construct subjective weights for all criteria and sub-criteria. Then, linguistic values characterized by triangular fuzzy numbers and trapezoidal fuzzy numbers are used to denote the evaluation values of all alternatives with respect to various subjective and objective criteria. Finally, the aggregation fuzzy assessment of different shipping companies is ranked to determine the best selection. Utilizing this fuzzy MCDM model, the decision-maker's fuzzy assessment and the trade-off between various evaluations criteria can be taken into account in the aggregation process, thus ensuring more effective and accurate decision-making.</code>

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>allnli</summary>

• Dataset: allnli at d482672
• Size: 557,850 training samples
• Columns: <code>anchor</code>, <code>positive</code>, and <code>negative</code>
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	<ul><li>min: 18 characters</li><li>mean: 34.88 characters</li><li>max: 193 characters</li></ul>	<ul><li>min: 15 characters</li><li>mean: 46.49 characters</li><li>max: 181 characters</li></ul>	<ul><li>min: 16 characters</li><li>mean: 50.47 characters</li><li>max: 204 characters</li></ul>

• Samples:

  anchor	positive	negative
  <code>A person on a horse jumps over a broken down airplane.</code>	<code>A person is outdoors, on a horse.</code>	<code>A person is at a diner, ordering an omelette.</code>
  <code>Children smiling and waving at camera</code>	<code>There are children present</code>	<code>The kids are frowning</code>
  <code>A boy is jumping on skateboard in the middle of a red bridge.</code>	<code>The boy does a skateboarding trick.</code>	<code>The boy skates down the sidewalk.</code>

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>paq</summary>

• Dataset: paq at 74601d8
• Size: 64,371,441 training samples
• Columns: <code>query</code> and <code>answer</code>
• Approximate statistics based on the first 1000 samples:

  	query	answer
  type	string	string
  details	<ul><li>min: 25 characters</li><li>mean: 50.56 characters</li><li>max: 104 characters</li></ul>	<ul><li>min: 509 characters</li><li>mean: 620.96 characters</li><li>max: 773 characters</li></ul>

• Samples:

  query	answer
  <code>in veetla visheshanga ganesh is the husband of</code>	<code>Veetla Visheshanga a song which reminds Ganga's memory. She is actually not Ganga but Gowri and her lover is the groom named Ganesh. When both were about to marry they were stopped by some goons because of which Gowri fell from the mountain but survived with injuries. Gopal who found the truth brought Ganesh to unite them. Gopal insists Gowri to marry Ganesh as both of them are lovers to which Gowri unwillingly accepts. But while Ganesh tries to tie the Mangal Sutra, Gowri stops him and she goes to Gopal saying that he may not need her but she needs him</code>
  <code>when did simon property group became a publicly traded company</code>	<code>of the S&P 100. Simon Property Group has been the subject of several lawsuits and investigations regarding civil rights and discrimination. Simon Property Group was formed in 1993 when the majority of the shopping center interests of Melvin Simon & Associates became a publicly traded company. Melvin Simon & Associates, owned by brothers Melvin Simon and Herbert Simon, was founded in 1960 in Indianapolis, Indiana, and had long been one of the top shopping center developers in the United States. In 1996, Simon DeBartolo Group was created when Simon Property merged with former rival DeBartolo Realty Corp. This was shortly</code>
  <code>what was the nationality of antoine faivre</code>	<code>Theosophy (Boehmian) below. "Theosophy": The scholar of esotericism Wouter Hanegraaff described Christian theosophy as "one of the major currents in the history of Western esotericism". Christian theosophy is an under-researched area; a general history of it has never been written. The French scholar Antoine Faivre had a specific interest in the theosophers and illuminists of the eighteenth and nineteenth centuries. He wrote his doctoral thesis on Karl von Eckartshausen and Christian theosophy. Scholars of esotericism have argued that Faivre's definition of Western esotericism relies on his own specialist focus on Christian theosophy, Renaissance Hermeticism, and Romantic "Naturphilosophie" and therefore creates an "ideal"</code>

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>trivia</summary>_qa

• Dataset: trivia_qa at a7c36e3
• Size: 73,346 training samples
• Columns: <code>query</code> and <code>answer</code>
• Approximate statistics based on the first 1000 samples:

  	query	answer
  type	string	string
  details	<ul><li>min: 21 characters</li><li>mean: 76.91 characters</li><li>max: 455 characters</li></ul>	<ul><li>min: 136 characters</li><li>mean: 3273.89 characters</li><li>max: 4096 characters</li></ul>

• Samples:

  query	answer
  <code>What type of rock is formed by the solidification of molten magma?</code>	<code>igneous rock - Dictionary Definition : Vocabulary.com igneous rock n rock formed by the solidification of molten magma Types: a rare type of peridotite that sometimes contains diamonds; found in South Africa and Siberia Type of: material consisting of the aggregate of minerals like those making up the Earth's crust Word Family Usage Examples Sign up, it's free! Whether you're a student, an educator, or a life-long learner, Vocabulary.com can put you on the path to systematic vocabulary improvement.</code>
  <code>Which river flows through the town of Shrewsbury?</code>	<code>River Severn
  <code>Which band's name was inspired by a novel by Herman Hesse?</code>	<code>23 Band Names Inspired by Literature :: Books :: Lists :: Paste 23 Band Names Inspired by Literature By Wyndham Wyeth

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>msmarco_10m</summary>

• Dataset: msmarco_10m at 8c5139a
• Size: 10,000,000 training samples
• Columns: <code>query</code>, <code>positive</code>, and <code>negative</code>
• Approximate statistics based on the first 1000 samples:

  	query	positive	negative
  type	string	string	string
  details	<ul><li>min: 10 characters</li><li>mean: 33.47 characters</li><li>max: 158 characters</li></ul>	<ul><li>min: 53 characters</li><li>mean: 353.76 characters</li><li>max: 948 characters</li></ul>	<ul><li>min: 67 characters</li><li>mean: 343.74 characters</li><li>max: 1063 characters</li></ul>

• Samples:

  query	positive	negative
  <code>what is ged equivalent</code>	<code>A GED is equivalent to a high school diploma however colleges do not look at it the same. A huge part of high school is commitment and dedication, therefore if you choose to drop out of high school you will probably not get accepted into any top colleges. However, you could always start at a community college and work your way up. Good Luck</code>	<code>If you are not far along in your Army career, education is an especially good way to boost your chances of promotion. Continuing your education at any level can earn you points. For completing your GED or bachelor's degree, you earn 10 points.</code>
  <code>foods that help with diverticulitis</code>	<code>Diet for Diverticulitis. Gradually you can ease back into a regular diet. Your doctor may advise you to start with low-fiber foods (white bread, meat, poultry, fish, eggs, and dairy products) before introducing high-fiber foods. Fiber softens and adds bulk to stools, helping them pass more easily through the colon.</code>	<code>During an attack of diverticulitis, your doctor may recommend a clear liquid diet or a low-fiber diet. This helps the area of infection to heal. Foods allowed on a clear-liquid diet include: 1 Plain water.iverticulitis occurs when small, bulging pouches (diverticula) in your colon become infected and inflamed — causing severe abdominal pain, nausea, and fever. The treatment of a diverticulitis attack will depend on the severity of the symptoms and whether this is your first attack.</code>
  <code>calories burned in turbo kick class</code>	<code>The American Council on Exercise did a study to find out how many calories are burned during a turbo kick class. The study looked at 15 women in a turbo kick class who weighed about 135 lbs and found that they burned between 6.45 to 8.3 calories a minute. This comes out to about 350 to 450 calories an hour.</code>	<code>Popular Calories Burned Searches: 1 Calories Burned For Intervals of run/walk or walk/jog: 6 mph or slower (> 10 minutes per mile) 2 Calories Burned For Walking: 3.5 mph (17 minutes per mile) 3 Calories Burned For Walking: 4 mph (15 minutes per mile) Calories Burned For Walking: 4.6 mph (13 minutes per mile)</code>

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>swim_ir</summary>

• Dataset: swim_ir at 834c20f
• Size: 501,538 training samples
• Columns: <code>query</code> and <code>text</code>
• Approximate statistics based on the first 1000 samples:

  	query	text
  type	string	string
  details	<ul><li>min: 0 characters</li><li>mean: 59.98 characters</li><li>max: 189 characters</li></ul>	<ul><li>min: 208 characters</li><li>mean: 525.9 characters</li><li>max: 2743 characters</li></ul>

• Samples:

  query	text
  <code>How many blocked kicks did Williams have in his second year at Bowling Green State University?</code>	<code>Williams accepted a football scholarship from Bowling Green State University, where he became one of the best special teams players in school history. As a redshirt freshman, he was a wide receiver on the scout team. The next year, he played mainly on special teams and had 3 blocked kicks.</code>
  <code>How many town councils are there in the metropolitan borough?</code>	<code>Horwich, Westhoughton and Blackrod are now constituted as civil parishes. There are three town councils in the metropolitan borough, Westhoughton Town Council, Horwich Town Council and Blackrod Town Council. The rest of the metropolitan borough, Bolton, Farnworth, Kearsley, Little Lever, and South Turton, have remained unparished areas since 1974.</code>
  <code>What is the name of the person selected to lead BART’s 296-member police force?</code>	<code>In 2009, the hiring of two independent organizations reviewed BART's policies and procedures in the process of overseeing the BART Police. The two independent firms investigated the matters of BART Police Shooting of Oscar Grant and were charged with making recommendations to the board. Ward Allen formulated and chaired BART's first Police Department Review Committee, and as a result, BART made sweeping changes on many security measures, as well as corrected and implemented several policies and procedures. BPD Review Committee has led to the re-training of all officers on use of force, diversity re-training and other issues. Ward Allen hired Kenton Rainey, the person selected to lead BART’s 296-member police force, to take command as Chief of Police.</code>

• Loss: <code>MatryoshkaLoss</code> with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

</details> <details><summary>pubmedqa</summary>

• Dataset: pubmedqa at a1ef0b5
• Size: 1,660 training samples
• Columns: <code>anchor</code>, <code>positive</code>, <code>negative_1</code>, <code>negative_2</code>, <code>negative_3</code>, <code>negative_4</code>, <code>negative_5</code>, <code>negative_6</code>, <code>negative_7</code>, <code>negative_8</code>, <code>negative_9</code>, <code>negative_10</code>, <code>negative_11</code>, <code>negative_12</code>, <code>negative_13</code>, <code>negative_14</code>, <code>negative_15</code>, <code>negative_16</code>, <code>negative_17</code>, <code>negative_18</code>, <code>negative_19</code>, and <code>negative_20</code>
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10	negative_11	negative_12	negative_13	negative_14	negative_15	negative_16	negative_17	negative_18	negative_19	negative_20
  type	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string
  details	<ul><li>min: 25 characters</li><li>mean: 94.06 characters</li><li>max: 213 characters</li></ul>	<ul><li>min: 5 characters</li><li>mean: 409.42 characters</li><li>max: 1582 characters</li></ul>	<ul><li>min: 5 characters</li><li>mean: 325.57 characters</li><li>max: 1300 characters</li></ul>	<ul><li>min: 17 characters</li><li>mean: 299.5 characters</li><li>max: 1352 characters</li></ul>	<ul><li>min: 21 characters</li><li>mean: 317.37 characters</li><li>max: 1590 characters</li></ul>	<ul><li>min: 13 characters</li><li>mean: 334.43 characters</li><li>max: 1536 characters</li></ul>	<ul><li>min: 13 characters</li><li>mean: 335.49 characters</li><li>max: 1247 characters</li></ul>	<ul><li>min: 13 characters</li><li>mean: 336.16 characters</li><li>max: 1383 characters</li></ul>	<ul><li>min: 14 characters</li><li>mean: 319.98 characters</li><li>max: 1501 characters</li></ul>	<ul><li>min: 16 characters</li><li>mean: 337.33 characters</li><li>max: 1493 characters</li></ul>	<ul><li>min: 15 characters</li><li>mean: 324.82 characters</li><li>max: 1058 characters</li></ul>	<ul><li>min: 13 characters</li><li>mean: 336.76 characters</li><li>max: 1457 characters</li></ul>	<ul><li>min: 10 characters</li><li>mean: 355.4 characters</li><li>max: 1748 characters</li></ul>	<ul><li>min: 13 characters</li><li>mean: 344.26 characters</li><li>max: 1705 characters</li></ul>	<ul><li>min: 16 characters</li><li>mean: 335.11 characters</li><li>max: 1593 characters</li></ul>	<ul><li>min: 18 characters</li><li>mean: 353.83 characters</li><li>max: 1374 characters</li></ul>	<ul><li>min: 13 characters</li><li>mean: 328.01 characters</li><li>max: 1755 characters</li></ul>	<ul><li>min: 12 characters</li><li>mean: 337.74 characters</li><li>max: 1579 characters</li></ul>	<ul><li>min: 16 characters</li><li>mean: 336.94 characters</li><li>max: 1325 characters</li></ul>	<ul><li>min: 15 characters</li><li>mean: 319.49 characters</li><li>max: 1410 characters</li></ul>	<ul><li>min: 12 characters</li><li>mean: 340.91 characters</li><li>max: 1680 characters</li></ul>	<ul><li>min: 20 characters</li><li>mean: 330.34 characters</li><li>max: 1509 characters</li></ul>

• Samples: | anchor | positive | negative_1 | negative_2 | negative_3 | negative_4 | negative_5 | negative_6 | negative_7 | negative_8 | negative_9 | negative_10 | negative_11 | negative_12 | negative_13 | negative_14 | negative_15 | negative_16 | negative_17 | negative_18 | negative_19 | negative_20 |

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