checkpoint-57/README.md

metadata

language:
  - en
license: apache-2.0
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:9432
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
base_model: sentence-transformers/all-MiniLM-L6-v2
widget:
  - source_sentence: >-
      Atherosclerosis and coronary heart disease are examples of what type of
      body system disease?
    sentences:
      - >-
        Diseases of the cardiovascular system are common and may be life
        threatening. Examples include atherosclerosis and coronary heart
        disease. A healthy lifestyle can reduce the risk of such diseases
        developing. This includes avoiding smoking, getting regular physical
        activity, and maintaining a healthy percent of body fat.
      - >-
        Osmosis Osmosis is the diffusion of water through a semipermeable
        membrane according to the concentration gradient of water across the
        membrane. Whereas diffusion transports material across membranes and
        within cells, osmosis transports only water across a membrane and the
        membrane limits the diffusion of solutes in the water. Osmosis is a
        special case of diffusion. Water, like other substances, moves from an
        area of higher concentration to one of lower concentration. Imagine a
        beaker with a semipermeable membrane, separating the two sides or halves
        (Figure 3.21). On both sides of the membrane, the water level is the
        same, but there are different concentrations on each side of a dissolved
        substance, or solute, that cannot cross the membrane. If the volume of
        the water is the same, but the concentrations of solute are different,
        then there are also different concentrations of water, the solvent, on
        either side of the membrane.
      - >-
        Circadian rhythms are regular changes in biology or behavior that occur
        in a 24-hour cycle. In humans, for example, blood pressure and body
        temperature change in a regular way throughout each 24-hour day. Animals
        may eat and drink at certain times of day as well. Humans have daily
        cycles of behavior, too. Most people start to get sleepy after dark and
        have a hard time sleeping when it is light outside. In many species,
        including humans, circadian rhythms are controlled by a tiny structure
        called the biological clock . This structure is located in a gland at
        the base of the brain. The biological clock sends signals to the body.
        The signals cause regular changes in behavior and body processes. The
        amount of light entering the eyes helps control the biological clock.
        The clock causes changes that repeat every 24 hours.
  - source_sentence: >-
      How does a cell's membrane keep extracellular materials from mixing with
      it's internal components?
    sentences:
      - >-
        We know that the Universe is expanding. Astronomers have wondered if it
        is expanding fast enough to escape the pull of gravity. Would the
        Universe just expand forever? If it could not escape the pull of
        gravity, would it someday start to contract? This means it would
        eventually get squeezed together in a big crunch. This is the opposite
        of the Big Bang.
      - >-
        Physical properties that do not depend on the amount of substance
        present are called intensive properties . Intensive properties do not
        change with changes of size, shape, or scale. Examples of intensive
        properties are as follows in the Table below .
      - >-
        CHAPTER REVIEW 3.1 The Cell Membrane The cell membrane provides a
        barrier around the cell, separating its internal components from the
        extracellular environment. It is composed of a phospholipid bilayer,
        with hydrophobic internal lipid “tails” and hydrophilic external
        phosphate “heads. ” Various membrane proteins are scattered throughout
        the bilayer, both inserted within it and attached to it peripherally.
        The cell membrane is selectively permeable, allowing only a limited
        number of materials to diffuse through its lipid bilayer. All materials
        that cross the membrane do so using passive (non energy-requiring) or
        active (energy-requiring) transport processes. During passive transport,
        materials move by simple diffusion or by facilitated diffusion through
        the membrane, down their concentration gradient. Water passes through
        the membrane in a diffusion process called osmosis. During active
        transport, energy is expended to assist material movement across the
        membrane in a direction against their concentration gradient. Active
        transport may take place with the help of protein pumps or through the
        use of vesicles.
  - source_sentence: An infection may be intracellular or extracellular, depending on this?
    sentences:
      - >-
        22.3 Magnetic Fields and Magnetic Field Lines • Magnetic fields can be
        pictorially represented by magnetic field lines, the properties of which
        are as follows: 1. The field is tangent to the magnetic field line.
        Field strength is proportional to the line density. Field lines cannot
        cross. Field lines are continuous loops.
      - >-
        Figure 24.13 The lifecycle of an ascomycete is characterized by the
        production of asci during the sexual phase. The haploid phase is the
        predominant phase of the life cycle.
      - >-
        Caffeine is an example of a psychoactive drug. It is found in coffee and
        many other products (see Table below ). Caffeine is a central nervous
        system stimulant . Like other stimulant drugs, it makes you feel more
        awake and alert. Other psychoactive drugs include alcohol, nicotine, and
        marijuana. Each has a different effect on the central nervous system.
        Alcohol, for example, is a depressant . It has the opposite effects of a
        stimulant like caffeine.
  - source_sentence: What does water treatment do to water?
    sentences:
      - >-
        Some solutes, such as sodium acetate, do not recrystallize easily.
        Suppose an exactly saturated solution of sodium acetate is prepared at
        50°C. As it cools back to room temperature, no crystals appear in the
        solution, even though the solubility of sodium acetate is lower at room
        temperature. A supersaturated solution is a solution that contains more
        than the maximum amount of solute that is capable of being dissolved at
        a given temperature. The recrystallization of the excess dissolved
        solute in a supersaturated solution can be initiated by the addition of
        a tiny crystal of solute, called a seed crystal. The seed crystal
        provides a nucleation site on which the excess dissolved crystals can
        begin to grow. Recrystallization from a supersaturated solution is
        typically very fast.
      - >-
        Figure 23.13, the esophagus runs a mainly straight route through the
        mediastinum of the thorax. To enter the abdomen, the esophagus
        penetrates the diaphragm through an opening called the esophageal
        hiatus.
      - >-
        Water treatment is a series of processes that remove unwanted substances
        from water. More processes are needed to purify water for drinking than
        for other uses.
  - source_sentence: >-
      There are only four possible bases that make up each dna nucleotide:
      adenine, guanine, thymine, and?
    sentences:
      - >-
        Metamorphism. This long word means “to change form. “ A rock undergoes
        metamorphism if it is exposed to extreme heat and pressure within the
        crust. With metamorphism , the rock does not melt all the way. The rock
        changes due to heat and pressure. A metamorphic rock may have a new
        mineral composition and/or texture.
      - >-
        Forest and Kim Starr (Flickr:Starr Environmental). Secondary succession
        occurs when nature reclaims areas formerly occupied by life . CC BY 2.0.
      - >-
        The only difference between each nucleotide is the identity of the base.
        There are only four possible bases that make up each DNA nucleotide:
        adenine (A), guanine (G), thymine (T), and cytosine (C).
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
model-index:
  - name: MNLP M3 Encoder SciQA
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 384
          type: dim_384
        metrics:
          - type: cosine_accuracy@1
            value: 0.6101048617731173
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8007626310772163
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8541468064823642
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9256434699714013
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6101048617731173
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2669208770257388
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.17082936129647283
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09256434699714014
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.6101048617731173
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8007626310772163
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8541468064823642
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9256434699714013
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7675175612283535
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7170116664396936
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7197084605820631
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 256
          type: dim_256
        metrics:
          - type: cosine_accuracy@1
            value: 0.5948522402287894
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.792183031458532
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8398474737845567
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9151572926596759
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.5948522402287894
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2640610104861773
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.16796949475691134
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09151572926596759
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.5948522402287894
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.792183031458532
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8398474737845567
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9151572926596759
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7548435122429773
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7035797509343749
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7070932589939358
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 192
          type: dim_192
        metrics:
          - type: cosine_accuracy@1
            value: 0.5910390848427073
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.7778836987607245
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8360343183984748
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9046711153479504
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.5910390848427073
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.25929456625357483
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.16720686367969495
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09046711153479504
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.5910390848427073
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.7778836987607245
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8360343183984748
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9046711153479504
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7477240665900656
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.6975449029309853
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7014228144337117
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 128
          type: dim_128
        metrics:
          - type: cosine_accuracy@1
            value: 0.567206863679695
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.7616777883698761
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8265014299332698
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.8903717826501429
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.567206863679695
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.253892596123292
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.16530028598665394
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.08903717826501431
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.567206863679695
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.7616777883698761
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8265014299332698
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.8903717826501429
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7273531110418706
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.6752920392815543
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.6794753898354032
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 96
          type: dim_96
        metrics:
          - type: cosine_accuracy@1
            value: 0.5529075309818875
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.7416587225929456
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8093422306959008
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.8741658722592945
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.5529075309818875
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.24721957419764853
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.1618684461391802
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.08741658722592945
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.5529075309818875
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.7416587225929456
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8093422306959008
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.8741658722592945
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7125237648315317
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.6608247461679306
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.6652525185575742
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 64
          type: dim_64
        metrics:
          - type: cosine_accuracy@1
            value: 0.5166825548141086
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.7054337464251669
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.7673975214489991
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.8369876072449952
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.5166825548141086
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.23514458214172226
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.1534795042897998
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.08369876072449953
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.5166825548141086
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.7054337464251669
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.7673975214489991
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.8369876072449952
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.6755921916053389
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.6240088822309986
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.629350282837756
            name: Cosine Map@100

MNLP M3 Encoder SciQA

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2 on the json dataset. It maps sentences & paragraphs to a 384-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: sentence-transformers/all-MiniLM-L6-v2
• Maximum Sequence Length: 256 tokens
• Output Dimensionality: 384 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • json
• 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': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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})
  (2): Normalize()
)

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("sentence_transformers_model_id")
# Run inference
sentences = [
    'There are only four possible bases that make up each dna nucleotide: adenine, guanine, thymine, and?',
    'The only difference between each nucleotide is the identity of the base. There are only four possible bases that make up each DNA nucleotide: adenine (A), guanine (G), thymine (T), and cytosine (C).',
    'Metamorphism. This long word means “to change form. “ A rock undergoes metamorphism if it is exposed to extreme heat and pressure within the crust. With metamorphism , the rock does not melt all the way. The rock changes due to heat and pressure. A metamorphic rock may have a new mineral composition and/or texture.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

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

Evaluation

Metrics

Information Retrieval

• Datasets: dim_384, dim_256, dim_192, dim_128, dim_96 and dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	dim_384	dim_256	dim_192	dim_128	dim_96	dim_64
cosine_accuracy@1	0.6101	0.5949	0.591	0.5672	0.5529	0.5167
cosine_accuracy@3	0.8008	0.7922	0.7779	0.7617	0.7417	0.7054
cosine_accuracy@5	0.8541	0.8398	0.836	0.8265	0.8093	0.7674
cosine_accuracy@10	0.9256	0.9152	0.9047	0.8904	0.8742	0.837
cosine_precision@1	0.6101	0.5949	0.591	0.5672	0.5529	0.5167
cosine_precision@3	0.2669	0.2641	0.2593	0.2539	0.2472	0.2351
cosine_precision@5	0.1708	0.168	0.1672	0.1653	0.1619	0.1535
cosine_precision@10	0.0926	0.0915	0.0905	0.089	0.0874	0.0837
cosine_recall@1	0.6101	0.5949	0.591	0.5672	0.5529	0.5167
cosine_recall@3	0.8008	0.7922	0.7779	0.7617	0.7417	0.7054
cosine_recall@5	0.8541	0.8398	0.836	0.8265	0.8093	0.7674
cosine_recall@10	0.9256	0.9152	0.9047	0.8904	0.8742	0.837
cosine_ndcg@10	0.7675	0.7548	0.7477	0.7274	0.7125	0.6756
cosine_mrr@10	0.717	0.7036	0.6975	0.6753	0.6608	0.624
cosine_map@100	0.7197	0.7071	0.7014	0.6795	0.6653	0.6294

Training Details

Training Dataset

json

• Dataset: json
• Size: 9,432 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 7 tokens mean: 18.15 tokens max: 60 tokens	min: 10 tokens mean: 94.56 tokens max: 256 tokens

• Samples:

  anchor	positive
  What is the term for atherosclerosis of arteries that supply the heart muscle?	Atherosclerosis of arteries that supply the heart muscle is called coronary heart disease . This disease may or may not have symptoms, such as chest pain. As the disease progresses, there is an increased risk of heart attack. A heart attack occurs when the blood supply to part of the heart muscle is blocked and cardiac muscle fibers die. Coronary heart disease is the leading cause of death of adults in the United States.
  What term describes a drug that has an effect on the central nervous system?	Caffeine is an example of a psychoactive drug. It is found in coffee and many other products (see Table below ). Caffeine is a central nervous system stimulant . Like other stimulant drugs, it makes you feel more awake and alert. Other psychoactive drugs include alcohol, nicotine, and marijuana. Each has a different effect on the central nervous system. Alcohol, for example, is a depressant . It has the opposite effects of a stimulant like caffeine.
  What scale is used to succinctly communicate the acidity or basicity of a solution?	The pH scale is used to succinctly communicate the acidity or basicity of a solution.

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          384,
          256,
          192,
          128,
          96,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 4
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: True
• tf32: True
• load_best_model_at_end: True
• optim: adamw_torch_fused
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• 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: 4
• max_steps: -1
• lr_scheduler_type: cosine
• 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: True
• 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: True
• 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}
• tp_size: 0
• 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_fused
• 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
• 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
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_384_cosine_ndcg@10	dim_256_cosine_ndcg@10	dim_192_cosine_ndcg@10	dim_128_cosine_ndcg@10	dim_96_cosine_ndcg@10	dim_64_cosine_ndcg@10
0.5424	10	22.4049	-	-	-	-	-	-
1.0	19	-	0.7424	0.7315	0.7263	0.7093	0.6919	0.6575
1.0542	20	16.6616	-	-	-	-	-	-
1.5966	30	16.8367	-	-	-	-	-	-
2.0	38	-	0.7612	0.7520	0.7431	0.7261	0.7097	0.6708
2.1085	40	12.8169	-	-	-	-	-	-
2.6508	50	13.7826	-	-	-	-	-	-
3.0	57	-	0.7675	0.7548	0.7477	0.7274	0.7125	0.6756

Framework Versions

• Python: 3.12.8
• Sentence Transformers: 3.4.1
• Transformers: 4.51.3
• PyTorch: 2.5.1+cu124
• Accelerate: 1.3.0
• Datasets: 3.6.0
• Tokenizers: 0.21.0

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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}