SetFit with sentence-transformers/paraphrase-mpnet-base-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-mpnet-base-v2 as the Sentence Transformer embedding model. A LogisticRegression instance is used for classification.

The model has been trained using an efficient few-shot learning technique that involves:

  1. Fine-tuning a Sentence Transformer with contrastive learning.
  2. Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

Model Sources

Model Labels

Label Examples
positive
  • 'Policy statement revised; IV antibiotic therapy is not medically necessary for uncomplicated cranial nerve palsy associated with Lyme disease and antibiotic-refractory Lyme arthritis\n7/19/2007: Reviewed and approved by MPAC\n7/10/2009: Policy reviewed, no changes\n12/15/2009: Coding Section revised with 2010 CPT4 and HCPCS revisions\n02/23/2011: Added the following to the policy statement: Determination of levels of the B lymphocyte chemoattractant CXCL13 for diagnosis or monitoring treatment is considered investigational. No changes to other policy statements. Removed deleted HCPCS codes J0530, J0540, and J0550 from the Code Reference section. 02/24/2012: Add the following policy statement: A single 2- to 4-week course of IV antibiotics may be considered medically necessary in patients with Lyme carditis, as evidenced by positive serologic findings (defined above) and associated with a high degree of atrioventricular block or a PR interval of greater than 0.3 second. Documentation of Lyme carditis may include PCR-based direct detection of B burgdorferi in the blood when results of serologic studies are equivocal.'
  • 'It previously stated that determination of levels of the B lymphocyte chemoattractant CXCL13 for diagnosis or monitoring treatment is considered investigational. Deleted outdated references from the Sources section. 11/28/2012: Policy reviewed; no changes. 03/10/2014: Policy reviewed; no changes to policy statement. Removed deleted HCPCS codes J0560, J0570, and J0580 from the Code Reference section.'
  • 'No changes to other policy statements. Removed deleted HCPCS codes J0530, J0540, and J0550 from the Code Reference section. 02/24/2012: Add the following policy statement: A single 2- to 4-week course of IV antibiotics may be considered medically necessary in patients with Lyme carditis, as evidenced by positive serologic findings (defined above) and associated with a high degree of atrioventricular block or a PR interval of greater than 0.3 second. Documentation of Lyme carditis may include PCR-based direct detection of B burgdorferi in the blood when results of serologic studies are equivocal. The last policy statement was revised to state that other diagnostic testing is considered investigational including but not limited to C6 peptide ELISA or determination of levels of the B lymphocyte chemoattractant CXCL13 for diagnosis or monitoring treatment.'
negative
  • "Fido the Stamp is quite an innovative concept, blending pet ownership with personalized mailing. Many dog owners likely delight in the idea of featuring their pets' images on parcels. This personalization could make mailing feel more intimate and fun for both senders and recipients. However, one might wonder if everyone would want their pet's face plastered on every package they send. \n\nIt’s plausible that some users might find it amusing to enhance the mundane task of mailing with a dash of pet-inspired personality. Yet, others might feel that featuring their dog on a stamp diminishes the seriousness of the parcels they are sending. Fido the Stamp could also lead"
  • "Friedreich's ataxia, which presents with symptoms such as gait disturbances, tremors, and speech difficulties, is associated with mutations in the FXN gene. This condition involves GAA trinucleotide repeat expansions. The FXN gene is located on chromosome 9, specifically at the locus 9q21.11. Thus, chromosome 9 is most commonly associated with Friedreich's ataxia."
  • 'The subdued July inflation figures indicate that consumer prices are not increasing rapidly, which usually boosts investor confidence in fixed-income securities like Treasuries. Likewise, soaring oil prices suggest that consumers may start cutting back on spending, leading to a cooling of the overall economy. This cooling effect can trigger a flight to safety among investors who prefer stable returns in uncertain times, causing an uptick in Treasury demand. Moreover, summer BBQs would probably be less popular if gas prices continue to rise, as people will likely prioritize their spending on essentials. However, a decrease in consumer spending could ironically lead to increased economic activity as consumers save money instead of spending it'

Evaluation

Metrics

Label Accuracy
all 1.0

Uses

Direct Use for Inference

First install the SetFit library:

pip install setfit

Then you can load this model and run inference.

from setfit import SetFitModel

# Download from the 🤗 Hub
model = SetFitModel.from_pretrained("ashercn97/code-y-v1")
# Run inference
preds = model("Given the symptoms described, the most likely karyotype for this 15-year-old boy is 47,XXY, which is characteristic of Klinefelter syndrome. The combination of decreased facial and pubic hair, gynecomastia, small testes, long extremities, and tall stature aligns with this chromosomal pattern. Klinefelter syndrome is caused by the presence of an extra X chromosome, leading to the 47,XXY karyotype.")

Training Details

Training Set Metrics

Training set Min Median Max
Word count 52 148.16 266
Label Training Sample Count
negative 9
positive 16

Training Hyperparameters

  • batch_size: (16, 16)
  • num_epochs: (4, 4)
  • max_steps: -1
  • sampling_strategy: oversampling
  • body_learning_rate: (2e-05, 1e-05)
  • head_learning_rate: 0.01
  • loss: CosineSimilarityLoss
  • distance_metric: cosine_distance
  • margin: 0.25
  • end_to_end: False
  • use_amp: False
  • warmup_proportion: 0.1
  • l2_weight: 0.01
  • seed: 42
  • eval_max_steps: -1
  • load_best_model_at_end: True

Training Results

Epoch Step Training Loss Validation Loss
0.0435 1 0.1928 -
1.0 23 - 0.0154
2.0 46 - 0.0023
2.1739 50 0.0214 -
3.0 69 - 0.0018
4.0 92 - 0.0015

Framework Versions

  • Python: 3.11.10
  • SetFit: 1.1.2
  • Sentence Transformers: 4.0.2
  • Transformers: 4.51.3
  • PyTorch: 2.4.1+cu124
  • Datasets: 3.5.0
  • Tokenizers: 0.21.1

Citation

BibTeX

@article{https://doi.org/10.48550/arxiv.2209.11055,
    doi = {10.48550/ARXIV.2209.11055},
    url = {https://arxiv.org/abs/2209.11055},
    author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren},
    keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
    title = {Efficient Few-Shot Learning Without Prompts},
    publisher = {arXiv},
    year = {2022},
    copyright = {Creative Commons Attribution 4.0 International}
}
Downloads last month
42
Safetensors
Model size
109M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ashercn97/code-y-v1

Finetuned
(294)
this model

Evaluation results