JoaoVitorr/food-classification-model-v3

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: 8 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
Lanche	'X-Tudo completo' 'X-Bacon artesanal' 'Hamburguer duplo'
Japonesa	'Barca de Sushi' 'Temaki de Salmão' 'Sashimi'
Brasileira	'Feijoada completa' 'Prato Feito (PF)' 'Marmita'
Pizza/Massa	'Pizza de Calabresa' 'Pizza Portuguesa' 'Pizza Marguerita'
Sobremesa	'Petit Gateau' 'Bolo de Chocolate' 'Torta de Limão'
Bebida	'Coca-Cola Zero' 'Guaraná' 'Suco de Laranja'
Petiscos	'Batata Frita' 'Batata Frita com queijo' 'Porção de Batata'
Árabe	'Esfiha de Carne' 'Esfiha de Queijo' 'Esfiha de Frango'

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("JoaoVitorr/food-classification-model-v3")
# Run inference
preds = model("Mocotó")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	2.0837	5

Label	Training Sample Count
Bebida	26
Brasileira	26
Japonesa	23
Lanche	30
Petiscos	27
Pizza/Massa	21
Sobremesa	30
Árabe	20

Training Hyperparameters

• batch_size: (32, 32)
• num_epochs: (5, 5)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 20
• body_learning_rate: (1e-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
0.0039	1	0.3079	-
0.1969	50	0.2609	-
0.3937	100	0.2381	-
0.5906	150	0.2351	-
0.7874	200	0.2165	-
0.9843	250	0.1902	-
1.1811	300	0.1445	-
1.3780	350	0.1209	-
1.5748	400	0.0992	-
1.7717	450	0.0771	-
1.9685	500	0.0594	-
2.1654	550	0.0486	-
2.3622	600	0.0372	-
2.5591	650	0.0332	-
2.7559	700	0.0269	-
2.9528	750	0.0204	-
3.1496	800	0.0163	-
3.3465	850	0.0137	-
3.5433	900	0.0112	-
3.7402	950	0.0121	-
3.9370	1000	0.0104	-
4.1339	1050	0.0087	-
4.3307	1100	0.0079	-
4.5276	1150	0.008	-
4.7244	1200	0.0075	-
4.9213	1250	0.0075	-

Framework Versions

• Python: 3.12.12
• SetFit: 1.1.3
• Sentence Transformers: 5.1.2
• Transformers: 4.57.1
• PyTorch: 2.8.0+cu126
• Datasets: 4.0.0
• Tokenizers: 0.22.1

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}
}