README.md

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:53851
  - loss:MultipleNegativesRankingLoss
base_model: BAAI/bge-base-en-v1.5
widget:
  - source_sentence: >-
      A certain junior class has 1000 students and a certain senior class has
      900 students. Among these students, there are 60 siblings pairs each
      consisting of 1 junior and 1 senior. If 1 student is to be selected at
      random from each class, what is the probability that the 2 students
      selected will be a sibling pair?
    sentences:
      - >-
        Let's see Pick 60/1000 first Then we can only pick 1 other pair from the
        800 So total will be 60 / 900 *1000 Simplify and you get 2/30000
      - >-
        To maximize number of hot dogs with 300$ Total number of hot dogs bought
        in 250-pack = 22.95*13 =298.35$ Amount remaining = 300 - 298.35 = 1.65$
        This amount is too less to buy any 8- pack . Greatest number of hot dogs
        one can buy with 300 $ = 250*13 = 3250
      - artificial leg
  - source_sentence: >-
      A stock trader originally bought 300 shares of stock from a company at a
      total cost of m dollars. If each share was sold at 80% above the original
      cost per share of stock, then interns of m for how many dollars was each
      share sold?
    sentences:
      - >-
        Let Cost of 300 shares be $ 3000 So, Cost of 1 shares be $ 10 =>m/300
        Selling price per share = (100+80)/100 * m/300 Or, Selling price per
        share = 9/5 * m/300 => 9m/1500
      - >-
        The prognostic value of p53 nuclear accumulation in gastric cancer is
        still unclear, as shown by the discordant results still reported in the
        literature. In this study, we evaluated the correlation between p53
        accumulation and long-term survival of patients resected for intestinal
        and diffuse-type gastric cancer. Eighty-three patients with carcinoma of
        the intestinal type and 53 patients with carcinoma of the diffuse type
        were included in the study. Immunohistochemical staining of the paraffin
        sections was performed by using monoclonal antibody DO1; cases were
        considered positive when nuclear immunostaining was observed in 10% or
        more of the tumor cells. Prognostic significance of different variables
        was investigated by univariate and multivariate analysis. p53 positivity
        was found in 51.8% of intestinal-type and 50.9% of diffuse-type cases.
        No significant correlation between the rate of p53 overexpression and
        age, sex, tumor location, tumor size, depth of invasion, lymph node
        involvement, distant metastases, and surgical radicality was found in
        the two groups of patients. A statistically significant difference in
        survival rate was observed between p53-negative and p53-positive cases
        in the intestinal type (P < .05), confirmed by multivariate analysis (P
        < .005; relative risk = 3.09). On the contrary, no correlation with
        survival was found in diffuse-type cases according to p53
        overexpression.
      - >-
        Many animal behaviors occur in a regular cycle. Two types of cyclic
        behaviors are circadian rhythms and migration.
  - source_sentence: >-
      Are lactate levels in severe malarial anaemia associated with
      haemozoin-containing neutrophils and low levels of IL-12?
    sentences:
      - >-
        Hyperlactataemia is often associated with a poor outcome in severe
        malaria in African children. To unravel the complex pathophysiology of
        this condition the relationship between plasma lactate levels, parasite
        density, pro- and anti-inflammatory cytokines, and haemozoin-containing
        leucocytes was studied in children with severe falciparum malarial
        anaemia. Twenty-six children with a primary diagnosis of severe malarial
        anaemia with any asexual Plasmodium falciparum parasite density and Hb <
        5 g/dL were studied and the association of plasma lactate levels and
        haemozoin-containing leucocytes, parasite density, pro- and
        anti-inflammatory cytokines was measured. The same associations were
        measured in non-severe malaria controls (N = 60). Parasite density was
        associated with lactate levels on admission (r = 0.56, P < 0.005).
        Moreover, haemozoin-containing neutrophils and IL-12 were strongly
        associated with plasma lactate levels, independently of parasite density
        (r = 0.60, P = 0.003 and r = -0.46, P = 0.02, respectively). These
        associations were not found in controls with uncomplicated malarial
        anaemia.
      - >-
        one of two female reproductive organs that produces eggs and secretes
        estrogen.
      - hydrogen
  - source_sentence: >-
      Does phosphatidylethanol mediate its effects on the vascular endothelial
      growth factor via HDL receptor in endothelial cells?
    sentences:
      - >-
        Patients having previous bariatric surgery are at risk for weight regain
        and return of co-morbidities. If an anatomic basis for the failure is
        identified, many surgeons advocate revision or conversion to a Roux-en-Y
        gastric bypass. The aim of this study was to determine whether
        revisional bariatric surgery leads to sufficient weight loss and
        co-morbidity remission. From 2005-2012, patients undergoing revision
        were entered into a prospectively maintained database. Perioperative
        outcomes, including complications, weight loss, and co-morbidity
        remission, were examined for all patients with a history of a previous
        vertical banded gastroplasty (VBG) or Roux-en-Y gastric bypass (RYGB).
        Twenty-two patients with a history of RYGB and 56 with a history of VBG
        were identified. Following the revisional procedure, the RYGB group
        experienced 35.8% excess weight loss (%EWL) and a 31.8% morbidity rate.
        For the VBG group, patients experienced a 46.2% %EWL from their weight
        before the revisional operation with a 51.8% morbidity rate.
        Co-morbidity remission rate was excellent. Diabetes (VBG:100%, RYGB:
        85.7%), gastroesophageal reflux disease (VBG: 94.4%, RYGB: 80%), and
        hypertension (VBG: 74.2%, RYGB:60%) demonstrated significant
        improvement.
      - >-
        Explanation: Let A, B, C represent their respective weights. Then, we
        have: A + B + C = (45 x 3) = 135 …. (i) A + B = (40 x 2) = 80 …. (ii) B
        + C = (44 x 2) = 88 ….(iii) Adding (ii) and (iii), we get: A + 2B + C =
        168 …. (iv) Subtracting (i) from (iv), we get : B = 33. B’s weight = 33
        kg.
      - >-
        Previous epidemiological studies have shown that light to moderate
        alcohol consumption has protective effects against coronary heart
        disease but the mechanisms of the beneficial effect of alcohol are not
        known. Ethanol may increase high density lipoprotein (HDL) cholesterol
        concentration, augment the reverse cholesterol transport, or regulate
        growth factors or adhesion molecules. To study whether qualitative
        changes in HDL phospholipids mediate part of the beneficial effects of
        alcohol on atherosclerosis by HDL receptor, we investigated whether
        phosphatidylethanol (PEth) in HDL particles affects the secretion of
        vascular endothelial growth factor (VEGF) by a human scavenger receptor
        CD36 and LIMPII analog-I (CLA-1)-mediated pathway. Human EA.hy 926
        endothelial cells were incubated in the presence of native HDL or
        PEth-HDL. VEGF concentration and CLA-1 protein expression were measured.
        Human CLA-1 receptor-mediated mechanisms in endothelial cells were
        studied using CLA-1 blocking antibody and protein kinase inhibitors.
        Phosphatidylethanol-containing HDL particles caused a 6-fold increase in
        the expression of CLA-1 in endothelial cells compared with the effect of
        native HDL. That emergent effect was mediated mainly through protein
        kinase C and p44/42 mitogen-activated protein kinase pathways. PEth
        increased the secretion of VEGF and that increase could be abolished by
        a CLA-1 blocking antibody.
  - source_sentence: >-
      Said to go hand-in-hand with science, what evolves as new materials,
      designs, and processes are invented?
    sentences:
      - >-
        Technology evolves as new materials, designs, and processes are
        invented.
      - Technological design constraints may be physical or social.
      - >-
        let x=44444444,then 44444445=x+1 88888885=2x-3 44444442=x-2 44444438=x-6
        44444444^2=x^2 then substitute it in equation (x+1)(2x-3)(x-2)+(x-6)/x^2
        ans is 2x-5 i.e 88888883
pipeline_tag: sentence-similarity
library_name: sentence-transformers

SentenceTransformer based on BAAI/bge-base-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-base-en-v1.5. It maps sentences & paragraphs to a 768-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: BAAI/bge-base-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity

Model Sources

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

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

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("danthepol/MNLP_M3_document_encoder")
# Run inference
sentences = [
    'Said to go hand-in-hand with science, what evolves as new materials, designs, and processes are invented?',
    'Technology evolves as new materials, designs, and processes are invented.',
    'Technological design constraints may be physical or social.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

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

Training Details

Training Dataset

Unnamed Dataset

• Size: 53,851 training samples
• Columns: sentence_0 and sentence_1
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1
  type	string	string
  details	min: 8 tokens mean: 31.16 tokens max: 143 tokens	min: 3 tokens mean: 160.39 tokens max: 512 tokens

• Samples:

  sentence_0	sentence_1
  For integers U and V, when U is divided by V, the remainder is odd. Which of the following must be true?	At least one of U and V is odd
  A mailman puts .05% of letters in the wrong mailbox. How many deliveries must he make to misdeliver 2 items?	Let the number of total deliveries be x Then, .05% of x=2 (5/100)*(1/100)*x=2 x=4000
  A certain ball team has an equal number of right- and left-handed players. On a certain day, two-thirds of the players were absent from practice. Of the players at practice that day, two-third were left handed. What is the ratio of the number of right-handed players who were not at practice that day to the number of lefthanded players who were not at practice?	Say the total number of players is 18, 9 right-handed and 9 left-handed. On a certain day, two-thirds of the players were absent from practice --> 12 absent and 6 present. Of the players at practice that day, one-third were left-handed --> 6*2/3=4 were left-handed and 2 right-handed. The number of right-handed players who were not at practice that day is 9-2=7. The number of left-handed players who were not at practice that days is 9-4=5. The ratio = 7/5.

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• 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: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1
• num_train_epochs: 3
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• 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: False
• 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}
• 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
• 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: batch_sampler
• multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss
0.2971	500	0.1286
0.5942	1000	0.0769
0.8913	1500	0.0682
1.1884	2000	0.0416
1.4854	2500	0.0369
1.7825	3000	0.0326
2.0796	3500	0.0331
2.3767	4000	0.0213
2.6738	4500	0.0211
2.9709	5000	0.0207

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: 3.2.0
• 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",
}

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}
}