oyvindbs/setfit-nb-sbert-v2-large-norec-sentiment-binary

SetFit with NbAiLab/nb-sbert-v2-large

This is a SetFit model that can be used for Text Classification. This SetFit model uses NbAiLab/nb-sbert-v2-large 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: NbAiLab/nb-sbert-v2-large
• 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
1	'Den kan skryte av firehjulsdrift , automatgir og stor motor .' '« Command & Conquer » har blitt selve symbolet på tradisjonelle strategispill , serien fnyser av moderne vrier og såkalte « nyvinninger » .' 'Skriveren benytter sublimeringsteknikk , i likhet med Dell Photoprinter 540 .'
0	'Ingen av merkets to bestselgende biler har dermed nådd topp 30-listen hittil i år .' '« Kameo » tar forholdsvis lang tid å gjennomføre , men mye av årsaken til det er bergene av frustrasjon man møter her og der .' 'Dommerne etterlyste råskap , mens VGs musikkjournalist Morten Ståle Nilsen mente opptredenen ikke var helt heldig .'

Evaluation

Metrics

Label	Accuracy
all	0.8919

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("Filmen er til tider nesten hypnotisk vakker .")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	19.035	68

Label	Training Sample Count
0	72
1	128

Training Hyperparameters

• batch_size: (30, 30)
• num_epochs: (5, 5)
• 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: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0014	1	0.2952	-
0.0689	50	0.2914	-
0.1377	100	0.1583	-
0.2066	150	0.0103	-
0.2755	200	0.0012	-
0.3444	250	0.0005	-
0.4132	300	0.0003	-
0.4821	350	0.0003	-
0.5510	400	0.0002	-
0.6198	450	0.0001	-
0.6887	500	0.0001	-
0.7576	550	0.0001	-
0.8264	600	0.0001	-
0.8953	650	0.0001	-
0.9642	700	0.0001	-
1.0331	750	0.0001	-
1.1019	800	0.0001	-
1.1708	850	0.0001	-
1.2397	900	0.0001	-
1.3085	950	0.0001	-
1.3774	1000	0.0001	-
1.4463	1050	0.0001	-
1.5152	1100	0.0001	-
1.5840	1150	0.0001	-
1.6529	1200	0.0001	-
1.7218	1250	0.0001	-
1.7906	1300	0.0001	-
1.8595	1350	0.0001	-
1.9284	1400	0.0	-
1.9972	1450	0.0	-
2.0661	1500	0.0	-
2.1350	1550	0.0	-
2.2039	1600	0.0	-
2.2727	1650	0.0	-
2.3416	1700	0.0	-
2.4105	1750	0.0	-
2.4793	1800	0.0	-
2.5482	1850	0.0	-
2.6171	1900	0.0	-
2.6860	1950	0.0	-
2.7548	2000	0.0	-
2.8237	2050	0.0	-
2.8926	2100	0.0	-
2.9614	2150	0.0	-
3.0303	2200	0.0	-
3.0992	2250	0.0	-
3.1680	2300	0.0	-
3.2369	2350	0.0	-
3.3058	2400	0.0	-
3.3747	2450	0.0	-
3.4435	2500	0.0	-
3.5124	2550	0.0	-
3.5813	2600	0.0	-
3.6501	2650	0.0	-
3.7190	2700	0.0	-
3.7879	2750	0.0	-
3.8567	2800	0.0	-
3.9256	2850	0.0	-
3.9945	2900	0.0	-
4.0634	2950	0.0	-
4.1322	3000	0.0	-
4.2011	3050	0.0	-
4.2700	3100	0.0	-
4.3388	3150	0.0	-
4.4077	3200	0.0	-
4.4766	3250	0.0004	-
4.5455	3300	0.0003	-
4.6143	3350	0.0	-
4.6832	3400	0.0	-
4.7521	3450	0.0	-
4.8209	3500	0.0006	-
4.8898	3550	0.0	-
4.9587	3600	0.0	-

Framework Versions

• Python: 3.11.15
• SetFit: 1.1.3
• Sentence Transformers: 3.3.1
• Transformers: 4.44.2
• PyTorch: 2.11.0
• Datasets: 4.8.5
• Tokenizers: 0.19.1

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}
}