JohanHeinsen/ENO_Runaway_Advertisement_classifier_2.0

SetFit with JohanHeinsen/Old_News_Segmentation_SBERT_V0.1

This is a SetFit model that can be used for Text Classification. This SetFit model uses JohanHeinsen/Old_News_Segmentation_SBERT_V0.1 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 Type: SetFit
• Sentence Transformer body: JohanHeinsen/Old_News_Segmentation_SBERT_V0.1
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 512 tokens
• Number of Classes: 2 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
0	'En meget brav gammel adelig Dame i Augsburg, har legeret 600,000 Gylden til et Pigeinstitues Oprettelse.' 'Efter indkommen Anmeldelse fra vedkommende Strandtoldbetjent er der løst af Havet inddrevet: paa Østeragger Strand 1 Oxhoved Viin mkt. J Feene paa Tolbøl Strand et Ditto Dito med samme Mærke, paa Hvidberg v. A. Strand 1 Ditto Dito mkt. DL. R. paa Ørum Strand 1 Ditto Dito mkt. 1 Pupaa Steenberg Strand 1 Ditto Dito mkt. NeEieren eller Eierne til fornævnte Oxhoveder Vine indkaldes herved sub poena præclusi et perpetui silentii med Aar og Dags Varsel at indfinde sig ved Amtet for at legitimere Eiendomsretten, hvorefter det indkommende Auctionsbeløb, med Fradrag af alle lovlige Udgifter, skal vorde Vedkommende udbetalt. Thisted Amthuus, den 24de August 1833. Faye.' 'Ved Tallotteriets 1212te Trækning i Altona den 12te April udkom følgende Nummere:'
1	'En Pige 15 Aar gammel, liden af Vext, navnlig Anne Marie, er den 25 May 1761. fra sine Forældre undvigt, og da hende en Arv er tilfalden, saa ombedes hun, eller hvo hende skulde forekomme, at formode hende at indfinde sig hos mig, boende i Nyeboder i Kiøbenhavn paa Elsdyrs-Længden i No. 18, som er hendes Fader, Christen Matros ved 4de Divisions 8de Compagnie.' 'At fra Kronborg Fæstnings Arbeide den 2 Oct. Sidst er undvigt uærlige Slave Hans Hansen, fød i Roeskilde, 42 Aar gl., liden af Vext, maadelig af Lemmer, blaae af Øine og bruun af Haar, det bekiendtgiøres herved til alle og enhvers Efterretning ligesom man og tillige vil have enhver anmodet at anholde denne for den offentlige Sikkerhed farlige Person, hvor som helst han skulde antræffes, og derefter henbringe ham til nærmeste Arresthuus til Bevaring, hvorfra han, naar saadant Commandant-skabet paa Kronborg tilmeldes, strax skal vorde afhentet, og de paa hans Anholdelse, Arrest og Forplegning anvendte Bekostninger, samt de sædvanlige Opbringerpenge bliver betalt, og tiener tillige til Underretning, at fornævnte Slave ved sin Undvigelse ei havde andet end bare Skiorte paa Kroppen, men Slave Buxer, Strømper og Skoe paa Benene, og en rund Hat paa Hovedet, og har desuden et stort Ar paa det ene Been fra en langvarig Beenskade.' 'Af Kongens Regiment har Mousqueteer Carl Sverling absenteret sig, samme var klæd i en graa Frakke, rød Manchesters Vest og Buxer, koparret af Ansigt, 23 Aar gl. 65, Tom. Høy; den som tager ham op, levere ham til Casernene imod Douceur efter Forordningen.'

Evaluation

Metrics

Label	Accuracy	F1	Precision	Recall
all	0.9990	0.9916	0.9833	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("setfit_model_id")
# Run inference
preds = model("En ganske nye Vand-Filtrum af Holms Fabrik i Kjøbenhavn, destillerende 50 Potter Vand om Dagen er tilkjøbs i Stokkemarke Præstegaard.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	5	88.9318	1999

Label	Training Sample Count
0	2093
1	149

Training Hyperparameters

• batch_size: (12, 12)
• num_epochs: (2, 2)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 12
• 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
• l2_weight: 0.01
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0002	1	0.5665	-
0.0112	50	0.4302	-
0.0223	100	0.3677	-
0.0335	150	0.1981	-
0.0446	200	0.0642	-
0.0558	250	0.0272	-
0.0669	300	0.0083	-
0.0781	350	0.0114	-
0.0892	400	0.0038	-
0.1004	450	0.0036	-
0.1115	500	0.0023	-
0.1227	550	0.005	-
0.1338	600	0.0031	-
0.1450	650	0.0011	-
0.1561	700	0.0038	-
0.1673	750	0.0001	-
0.1784	800	0.0005	-
0.1896	850	0.0019	-
0.2007	900	0.0016	-
0.2119	950	0.0001	-
0.2230	1000	0.0014	-
0.2342	1050	0.0022	-
0.2453	1100	0.0021	-
0.2565	1150	0.0018	-
0.2676	1200	0.0002	-
0.2788	1250	0.0	-
0.2899	1300	0.0019	-
0.3011	1350	0.0	-
0.3122	1400	0.0	-
0.3234	1450	0.0036	-
0.3345	1500	0.0	-
0.3457	1550	0.0	-
0.3568	1600	0.0	-
0.3680	1650	0.0	-
0.3791	1700	0.0	-
0.3903	1750	0.0018	-
0.4014	1800	0.0001	-
0.4126	1850	0.0017	-
0.4237	1900	0.0	-
0.4349	1950	0.0	-
0.4460	2000	0.0	-
0.4572	2050	0.0035	-
0.4683	2100	0.0034	-
0.4795	2150	0.0036	-
0.4906	2200	0.0017	-
0.5018	2250	0.0056	-
0.5129	2300	0.0006	-
0.5241	2350	0.0	-
0.5352	2400	0.0	-
0.5464	2450	0.0	-
0.5575	2500	0.0016	-
0.5687	2550	0.0014	-
0.5798	2600	0.0	-
0.5910	2650	0.0012	-
0.6021	2700	0.0001	-
0.6133	2750	0.0	-
0.6244	2800	0.0	-
0.6356	2850	0.0	-
0.6467	2900	0.0	-
0.6579	2950	0.0	-
0.6690	3000	0.0016	-
0.6802	3050	0.0	-
0.6913	3100	0.0	-
0.7025	3150	0.0	-
0.7136	3200	0.0017	-
0.7248	3250	0.0012	-
0.7360	3300	0.0002	-
0.7471	3350	0.0	-
0.7583	3400	0.0	-
0.7694	3450	0.0	-
0.7806	3500	0.0	-
0.7917	3550	0.0	-
0.8029	3600	0.0	-
0.8140	3650	0.0	-
0.8252	3700	0.0	-
0.8363	3750	0.0	-
0.8475	3800	0.0	-
0.8586	3850	0.0	-
0.8698	3900	0.0	-
0.8809	3950	0.0	-
0.8921	4000	0.0	-
0.9032	4050	0.0	-
0.9144	4100	0.0	-
0.9255	4150	0.0	-
0.9367	4200	0.0	-
0.9478	4250	0.0	-
0.9590	4300	0.0	-
0.9701	4350	0.0	-
0.9813	4400	0.0	-
0.9924	4450	0.0	-
1.0036	4500	0.0	-
1.0147	4550	0.0	-
1.0259	4600	0.0	-
1.0370	4650	0.0	-
1.0482	4700	0.0	-
1.0593	4750	0.0	-
1.0705	4800	0.0	-
1.0816	4850	0.0	-
1.0928	4900	0.0	-
1.1039	4950	0.0	-
1.1151	5000	0.0	-
1.1262	5050	0.0	-
1.1374	5100	0.0	-
1.1485	5150	0.0	-
1.1597	5200	0.0	-
1.1708	5250	0.0	-
1.1820	5300	0.0	-
1.1931	5350	0.0	-
1.2043	5400	0.0	-
1.2154	5450	0.0	-
1.2266	5500	0.0	-
1.2377	5550	0.0	-
1.2489	5600	0.0	-
1.2600	5650	0.0	-
1.2712	5700	0.0	-
1.2823	5750	0.0	-
1.2935	5800	0.0	-
1.3046	5850	0.0	-
1.3158	5900	0.0	-
1.3269	5950	0.0	-
1.3381	6000	0.0	-
1.3492	6050	0.0	-
1.3604	6100	0.0	-
1.3715	6150	0.0	-
1.3827	6200	0.0	-
1.3938	6250	0.0	-
1.4050	6300	0.0	-
1.4161	6350	0.0	-
1.4273	6400	0.0	-
1.4384	6450	0.0	-
1.4496	6500	0.0	-
1.4607	6550	0.0	-
1.4719	6600	0.0	-
1.4831	6650	0.0	-
1.4942	6700	0.0	-
1.5054	6750	0.0	-
1.5165	6800	0.0	-
1.5277	6850	0.0	-
1.5388	6900	0.0	-
1.5500	6950	0.0	-
1.5611	7000	0.0	-
1.5723	7050	0.0	-
1.5834	7100	0.0	-
1.5946	7150	0.0	-
1.6057	7200	0.0	-
1.6169	7250	0.0	-
1.6280	7300	0.0	-
1.6392	7350	0.0	-
1.6503	7400	0.0	-
1.6615	7450	0.0	-
1.6726	7500	0.0	-
1.6838	7550	0.0	-
1.6949	7600	0.0	-
1.7061	7650	0.0	-
1.7172	7700	0.0	-
1.7284	7750	0.0	-
1.7395	7800	0.0	-
1.7507	7850	0.0	-
1.7618	7900	0.0	-
1.7730	7950	0.0	-
1.7841	8000	0.0	-
1.7953	8050	0.0	-
1.8064	8100	0.0	-
1.8176	8150	0.0	-
1.8287	8200	0.0	-
1.8399	8250	0.0	-
1.8510	8300	0.0	-
1.8622	8350	0.0	-
1.8733	8400	0.0	-
1.8845	8450	0.0	-
1.8956	8500	0.0	-
1.9068	8550	0.0	-
1.9179	8600	0.0	-
1.9291	8650	0.0	-
1.9402	8700	0.0	-
1.9514	8750	0.0	-
1.9625	8800	0.0	-
1.9737	8850	0.0	-
1.9848	8900	0.0	-
1.9960	8950	0.0	-

Framework Versions

• Python: 3.11.12
• SetFit: 1.1.3
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.7.0
• Datasets: 2.19.2
• 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}
}