johnpaulbin/toki-pona-classifier-v2

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
toki pona	'ona li toki "toki" tawa meli.' 'toki li pona tawa mi.' 'mi toki e ni tawa ona: "o kama tawa tomo mi."'
other	'No te puedo creer el grado de precisión 🤣' 'i can’t deny i’m invested in the aspect of things :’)' "I'm live on #twitch, and speedrunning EarthBound!"

Evaluation

Metrics

Label	Accuracy
all	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("johnpaulbin/toki-pona-classifier-v2")
# Run inference
preds = model(["Hello!", "toki!"])

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	10.5705	61

Label	Training Sample Count
other	2035
toki pona	2000

Training Hyperparameters

• batch_size: (12, 12)
• num_epochs: (2, 2)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 1
• 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
0.0015	1	0.3252	-
0.0743	50	0.2704	-
0.1486	100	0.2257	-
0.2229	150	0.0567	-
0.2972	200	0.0063	-
0.3715	250	0.0015	-
0.4458	300	0.0034	-
0.5201	350	0.0026	-
0.5944	400	0.0036	-
0.6686	450	0.0005	-
0.7429	500	0.0021	-
0.8172	550	0.0021	-
0.8915	600	0.0003	-
0.9658	650	0.0002	-
1.0401	700	0.0002	-
1.1144	750	0.0018	-
1.1887	800	0.0003	-
1.2630	850	0.0002	-
1.3373	900	0.0001	-
1.4116	950	0.0015	-
1.4859	1000	0.0004	-
1.5602	1050	0.0001	-
1.6345	1100	0.0001	-
1.7088	1150	0.0019	-
1.7831	1200	0.0001	-
1.8574	1250	0.0001	-
1.9316	1300	0.0001	-

Framework Versions

• Python: 3.10.12
• SetFit: 1.1.0
• Sentence Transformers: 3.2.1
• Transformers: 4.42.2
• PyTorch: 2.5.1+cu121
• Datasets: 3.1.0
• 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}
}