saraestevez/setfit-minilm-bank-tweets-processed-400

SetFit with sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-multilingual-MiniLM-L12-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 Type: SetFit
• Sentence Transformer body: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 128 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
discard	'Marcos informa que se puede realizar el pago de productos de BBVA a través de la Línea BBVA, cajeros automáticos, practicajas, ventanilla de sucursal o diversos comercios.' 'Se ha celebrado una reunión de alto nivel en 2024 para concretar proyectos de inversión, incluyendo la cooperación con BBVA para la construcción de un portadrones y en el ámbito turístico.' 'Diversificar es clave para alcanzar nuestros objetivos en inversiones y en la vida, descubre cómo tus decisiones financieras pueden impactar tu vida personal en este artículo.'
relevant	'La persona recibió un correo idéntico al que le explicaron que es una técnica de estafa que simula enviarlo desde su propia cuenta.' 'La cancelación de la cuenta se ha demorado un mes y al solicitar 200 euros para un viaje, me han cobrado 9 euros de comisión.' 'El Santander logró récords en beneficios y comisiones a los desfavorecidos bajo el ministerio del consagrado en Consumo, mientras se obsesionan con la apariencia y carecen de dignidad y principios.'

Evaluation

Metrics

Label	Accuracy
all	0.8029

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("saraestevez/setfit-minilm-bank-tweets-processed-400")
# Run inference
preds = model("La app de BBVA está caída, pero se pide paciencia para los depósitos de mañana.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	21.6612	44

Label	Training Sample Count
discard	400
relevant	400

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (1, 1)
• 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
0.0005	1	0.3197	-
0.025	50	0.2199	-
0.05	100	0.2876	-
0.075	150	0.2568	-
0.1	200	0.196	-
0.125	250	0.15	-
0.15	300	0.1475	-
0.175	350	0.081	-
0.2	400	0.0441	-
0.225	450	0.0228	-
0.25	500	0.0017	-
0.275	550	0.0083	-
0.3	600	0.002	-
0.325	650	0.0013	-
0.35	700	0.0011	-
0.375	750	0.0014	-
0.4	800	0.0004	-
0.425	850	0.0001	-
0.45	900	0.0118	-
0.475	950	0.0002	-
0.5	1000	0.0012	-
0.525	1050	0.0003	-
0.55	1100	0.0001	-
0.575	1150	0.0003	-
0.6	1200	0.0001	-
0.625	1250	0.0001	-
0.65	1300	0.0001	-
0.675	1350	0.0002	-
0.7	1400	0.0197	-
0.725	1450	0.0002	-
0.75	1500	0.0002	-
0.775	1550	0.0001	-
0.8	1600	0.0004	-
0.825	1650	0.0001	-
0.85	1700	0.0001	-
0.875	1750	0.0001	-
0.9	1800	0.0001	-
0.925	1850	0.0001	-
0.95	1900	0.0158	-
0.975	1950	0.0001	-
1.0	2000	0.0001	-

Framework Versions

• Python: 3.11.0rc1
• SetFit: 1.0.3
• Sentence Transformers: 2.7.0
• Transformers: 4.39.0
• PyTorch: 2.3.1+cu121
• Datasets: 2.19.1
• Tokenizers: 0.15.2

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}
}