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 SetFitHead instance is used for classification.

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

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

Model Details

Model Description

Model Type: SetFit
Sentence Transformer body: sentence-transformers/paraphrase-mpnet-base-v2
Classification head: a SetFitHead instance
Maximum Sequence Length: 512 tokens
Number of Classes: 7 classes

Model Sources

Repository: SetFit on GitHub
Paper: Efficient Few-Shot Learning Without Prompts
Blogpost: SetFit: Efficient Few-Shot Learning Without Prompts

Model Labels

Label	Examples
6	'If you were especially helpful in a corrupt scheme you received not just cash in a bag , but equity . ' 'Two American companies reached deals for fields auctioned in June . ' 'Let me prove it , Phil . '
2	'This building shook like hell and it kept getting stronger . ' 'Now you could ask me , why should the user mind about MathML ? ' 'The report and a casebook of initiatives will be published in 1996 and provide the backdrop for a conference to be staged in Autumn , 1996 . '
3	'The tumor , he suggested , developed when the second , normal copy also was damaged . ' 'Proper English bells are started off in `rounds ,` from the highest-pitched bell to the lowest -- a simple descending scale using , in larger churches , as many as 12 bells . ' 'Treatment should be delayed or discontinued , or the dose reduced , in patients whose blood counts are abnormal or who have certain other side effects . '
5	'Schools that are structured in this way produce students with higher morale and superior academic performance . ' 'I got home , let the dogs into the house and noticed some sounds above my head , as if someone were walking on the roof , or upstairs . ' 'Give me your address . '
0	'-- Most important of all , schools should have principals with a large measure of authority over the faculty , the curriculum , and all matters of student discipline . ' 'For months the Johns Hopkins researchers , using gene probes , experimentally crawled down the length of chromosome 17 , looking for the smallest common bit of genetic material lost in all tumor cells . ' 'It explains how the Committee for Medicinal Products for Human Use ( CHMP ) assessed the studies performed , to reach their recommendations on how to use the medicine . '
4	'In 2005 , the fear of invasion of the national territory by hordes of Polish plumbers was felt both on the Left and on the Right . ' 'Cerenia contains the active substance maropitant and is available as tablet or as solution for injection . ' 'The second quarter was more of the same , but the Alavan team opted for the inside game of Barac and the work of Eliyahu , who was greeted with whistles and applause at his return home , to continue increasing their lead by half-time ( 34-43 ) . '
1	'`The sound of bells is a net to draw people into the church ,` he says . ' '`Progressive education` ( as it was once called ) is far more interesting and agreeable to teachers than is disciplined instruction . ' "The defense lawyers also claim , for example , that Mr. Hayes may have been prejudiced when Judge Blue declined to allow them to test potential jurors ' reactions by showing them grisly crime-scene photographs during jury selection . "

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("HelgeKn/SemEval-multi-class-v1-10")
# Run inference
preds = model("`` But why pay her bills ? ")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	5	25.8286	75

Label	Training Sample Count
0	10
1	10
2	10
3	10
4	10
5	10
6	10

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (4, 4)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 20
body_learning_rate: (2e-05, 2e-05)
head_learning_rate: 2e-05
loss: CosineSimilarityLoss
distance_metric: cosine_distance
margin: 0.25
end_to_end: False
use_amp: False
warmup_proportion: 0.1
seed: 42
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0057	1	0.2314	-
0.2857	50	0.218	-
0.5714	100	0.1161	-
0.8571	150	0.0559	-
1.1429	200	0.0087	-
1.4286	250	0.0029	-
1.7143	300	0.001	-
2.0	350	0.0006	-
2.2857	400	0.0011	-
2.5714	450	0.0009	-
2.8571	500	0.0005	-
3.1429	550	0.0006	-
3.4286	600	0.0004	-
3.7143	650	0.0003	-
4.0	700	0.0005	-

Framework Versions

Python: 3.9.13
SetFit: 1.0.1
Sentence Transformers: 2.2.2
Transformers: 4.36.0
PyTorch: 2.1.1+cpu
Datasets: 2.15.0
Tokenizers: 0.15.0

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}
}

HelgeKn
/

SemEval-multi-class-v1-10