SentenceTransformer based on microsoft/mpnet-base

This is a sentence-transformers model finetuned from microsoft/mpnet-base on the json dataset. 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: microsoft/mpnet-base
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • json

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': False, 'architecture': 'MPNetModel'})
  (1): Pooling({'word_embedding_dimension': 768, '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("sahithkumar7/final-mpnet-base-fullfinetuned-epoch5")
# Run inference
sentences = [
    'What is the number of genes obtained from comparing control and LIPUS-stimulated samples?',
    'Differentially expressed genes (DEGs) were obtained\nbetween control and LIPUS-stimulated samples using\nan adjusted P<0.05 and|log2FC| > 1 as cutoffs to define\nstatistically significant differential expression. 676 genes\nwere obtained from which 578 were upregulated when\nstimulated with LIPUS and 98 genes were subregulated\n(Supp. Figure 1). To further understand the functions\nand pathways associated with the differentially expressed\ngenes (DEG), Gene Ontology (GO) and Kyoto Encyclo-\npedia of Genes and Genomes (KEGG) analyses were con-\nducted using the DAVID database [37, 38].',
    'independent studies have shown a raising trend in both cancer incidence [2] and a high-salt\ndietary lifestyle [7], there is no direct correlation between dietary salt intake and breast\ncancer. Interestingly, in the human body, certain organs such as the skin and lymph nodes\nhave a natural tendency to accumulate salt [8]. Although unknown, the pathophysiological\nsignificance of this selective accumulation of sodium in certain organs and solid tumors is\nan area of intense research.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000,  0.6568, -0.0120],
#         [ 0.6568,  1.0000, -0.0343],
#         [-0.0120, -0.0343,  1.0000]])

Evaluation

Metrics

Triplet

• Datasets: initial_test and final_test
• Evaluated with TripletEvaluator

Training Details

Training Dataset

json

• Dataset: json
• Size: 800 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 800 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 7 tokensmean: 16.79 tokensmax: 39 tokens	min: 39 tokensmean: 117.74 tokensmax: 265 tokens	min: 40 tokensmean: 116.14 tokensmax: 356 tokens

• Samples:

  anchor	positive	negative
  What is the limitation of FBG-based sensors in tactile feedback?	Furthermore, FBG-based 3-axis tactile sensors have beenproposed for a more comprehensive haptic perception toolin surgeries (Figure 1D) (16). Five optical fibers mergedwith FBG sensors are suspended in a deformable mediumand measure the compression or tension of the tissue as thesensors are pressed against it, returning a _ surfacereaction map. While FBG-based sensors are small, flexible, andsensitive, there are several challenges that need to beaddressed for optimal performance for tactile feedback. Thesesensors are temperature sensitive, requiring temperature	141]. Therefore, it is not known to what extent sparedaxons are remyelinated by transplanted Schwann cells,nor is the contribution of this myelin to functional im-provements proven. Transplantation of Schwann cellsincapable of producing myelin, such as cells derivedfrom trembler (Pmp22Tr) mutant mice, may be usefulin establishing a causal relationship between myelin re-generation and functional improvements. Several MSCtransplantations demonstrate an increase of myelin re-tention and the number of myelinated axons in the le-sion site during a chronic post-injury period [57]. Thus,
  What are the advantages of strain elastography?	frontiersin.org--- Page 8 ---Kumar et al.TABLE 2 Modalities of ultrasound elastography.ModalityStrain elastographyExcitationApplied manual compression (38)AdvantagesNo additional specialized equipmentrequired (40)10.3389/fmedt.2023.1238129LimitationsQualitative measurements (39)Internal physiological mechanism (42)Simple low-cost design (40)Applied compression is operator-dependent (51)More commonly used (52)High inter-observer variability (51)coustic radiation force impulse Acoustic radiation force (43)(ARFI) imagingImage beyond slip boundaries (45)	Publisher’s Note: MDPI stays neutralwith regard to jurisdictional claims inpublished maps and institutional afil-iations.onon)Copyright: © 2021 by the author.Licensee MDPI, Basel, Switzerland.This article is an open access articledistributed under the terms andconditions of the Creative CommonsAttribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).Joan and Sanford I. Weill Department of Medicine, Weill Cornell Medical College, 525 East 68th Street,Room M-522, Box 130, New York, NY 10065, USA; str2020@med.cornell.edu or Stefan.Ryter@proterris.com
  What is the material used for the substrate in a piezoelectric element?	gain for biomedical applications.frontiersin.org--- Page 9 ---Kumar et al.>[PMUT ]Electrode: Voltage Electrode2© piezoelectric elementso—: OSi02©) silicon substrateB [ CMUT ]AC DCmembrane—————vacuuminsulatorsubstrate= groundFIGURE 3	HistopatholoCytology Total, n (%) Benign, n (%) P ey Cancer, n (%)FA 2 (15.4%) FTC 2 (25%)0 GD (7.7%) PTC 6 (75%)I 21 (4.0%) NG 9 (69.2%)Other diagnosis (7.7%)FA 15 (9.9%) FIC 4 (14.3%)FT-UMP (0.7%) MTC 3 (10.7%)GD (0.7%) PTC 21 (75%)Il 180 (34.5%) OA (0.7%)LT (0.7%)NG 130 (85.5%)NIFTP 2 (1.3%)FA 14 (23.7%) FIC 7 (28.0%)FI-UMP 2 (3.4%) OTC 1 (4.0%)OA (1.7%) PTC 17 (68.0%)Il 84 (16.1%) LT 3 (5.1%)NG 35 (59.3%)NIFTP 2 (3.4%)WDT-UMP 2 (3.4%)FA 15 (26.3%) OTC 1 (7.7%)FT-UMP 5 (8.8%) PTC 12 (92.3%)OA 13 (22.8%)IV 70 (13.4%) LT 2 (3.5%)NG 18 (31.6%)NIFTP 2 (3.5%)

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

json

• Dataset: json
• Size: 200 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 200 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 7 tokensmean: 17.14 tokensmax: 35 tokens	min: 40 tokensmean: 121.3 tokensmax: 356 tokens	min: 45 tokensmean: 119.75 tokensmax: 356 tokens

• Samples:

  anchor	positive	negative
  What can differentiate into a very wide variety of tissues?	lead to decreased rates of graft-versus-host disease. Theyalso can differentiate into a very wide variety of tissues. Forexample, when compared with bone marrow stem cells ormobilized peripheral blood, umbilical cord blood stem cellshave a greater repopulating ability.5° Cord blood derivedCD34+ cells have very potent hematopoietic abilities, andthis is attributed to the immaturity of the stem cells rela-tive to adult derived cells. Studies have been done that an-alyze long term survival of children with hematologic dis-orders who were transplanted with umbilical cord blood	metabolic regulation may affect the function of more than one organelle. Therefore, if themiR-17-92 regulatory cluster can perturb genes related to mitochondrial metabolic function,it could be also related, in some way, to genes involved in lysosomal metabolic function.Lysosomes are intracellular organelles that, in form of small vesicles, participate inseveral cellular functions, mainly digestion, but also vesicle trafficking, autophagy, nutrientsensing, cellular growth, signaling [85], and even enzyme secretion. The membrane-bound
  What are the two most common types of pluripotent stem cells?	III]. AMNIOTIC CELLS AS A SOURCE FOR STEMCELLSHistorically, the two most common types of pluripotentstem cells include embryonic stem cells (ESCs) and inducedpluripotent stem cells (iPSCs).35 However, despite the manyresearch efforts to improve ESC and iPSC technologies,there are still enormous clinical challenges.°> Two signif-icant issues posed by ESC and iPSC technologies includelow survival rate of transplanted cells and tumorigenicity.°>Recently, researchers have isolated pluripotent stem cells	Explanation: criterion 6 indicates a positive diagnosis only within the DC VI grouprelative to all other categories. Criterion 5 indicates a positive diagnosis within the DCs VIand V relative to all other categories.The highest positive predictive value (PPV) confirming malignancy through histopatho-logical examination for criterion 6 was 0.93, and for criterion 5, it was 0.92. For the subsequentcriteria, the PPVs were as follows: criterion 4—0.66; criterion 3—0.55; criterion 2—0.40.
  What percentage of stem cells are present in bone marrow?	ing 30% in some tissues.43-45 This is a significant differencefrom the .0001-.0002% stem cells present in bone marrow.43Given this difference in stem cell concentration betweenthe sources, there will be more ADSCs per sample of WAT	migration of bCSCs. This finding raises the possibil-ity that LIPUS may decrease the ability of these cells toinvade adjacent tissues and start the process of metasta-ses. These results also suggested that some of the changesinduced by LIPUS take longer to be detected in this typeof 2D migration model, possible due to changes in geneexpression pattern. To further study this hypothesis, weperformed a Transwell invasion assay. The data revealeda reduced number of cells crossing the membrane afterLIPUS stimulation, indicating that therapeutic LIPUS

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 5
• warmup_ratio: 0.1
• fp16: True
• batch_sampler: no_duplicates

All Hyperparameters

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• 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: 1.0
• num_train_epochs: 5
• 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: False
• fp16: True
• 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
• 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
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Framework Versions

• Python: 3.11.13
• Sentence Transformers: 5.0.0
• Transformers: 4.52.4
• PyTorch: 2.6.0+cu124
• Accelerate: 1.8.1
• Datasets: 3.6.0
• Tokenizers: 0.21.2

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}
}