johnpaulbin/toxicity-setfit-3-large

SetFit with jhu-clsp/mmBERT-base

This is a SetFit model that can be used for Text Classification. This SetFit model uses jhu-clsp/mmBERT-base 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: jhu-clsp/mmBERT-base
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 8192 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

Evaluation

Metrics

Label	Accuracy
all	0.9878

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/toxicity-setfit-3-large")
# Run inference
preds = model("just here")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	4.8969	81

Label	Training Sample Count
not toxic	9846
toxic	7132

Training Hyperparameters

• batch_size: (128, 128)
• num_epochs: (1, 1)
• max_steps: -1
• sampling_strategy: num_iterations
• num_iterations: 8
• 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: True
• warmup_proportion: 0.1
• l2_weight: 0.01
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0005	1	0.3202	-
0.0236	50	0.2709	-
0.0471	100	0.1783	-
0.0707	150	0.0719	-
0.0942	200	0.0453	-
0.1178	250	0.0266	-
0.1413	300	0.0175	-
0.1649	350	0.0158	-
0.1884	400	0.0102	-
0.2120	450	0.0075	-
0.2355	500	0.0072	-
0.2591	550	0.0064	-
0.2826	600	0.0041	-
0.3062	650	0.0028	-
0.3297	700	0.003	-
0.3533	750	0.0029	-
0.3768	800	0.0026	-
0.4004	850	0.0028	-
0.4239	900	0.0026	-
0.4475	950	0.0023	-
0.4710	1000	0.002	-
0.4946	1050	0.0009	-
0.5181	1100	0.0018	-
0.5417	1150	0.0008	-
0.5652	1200	0.0011	-
0.5888	1250	0.0019	-
0.6123	1300	0.001	-
0.6359	1350	0.0005	-
0.6594	1400	0.0012	-
0.6830	1450	0.0013	-
0.7065	1500	0.0003	-
0.7301	1550	0.0008	-
0.7537	1600	0.001	-
0.7772	1650	0.0009	-
0.8008	1700	0.001	-
0.8243	1750	0.0007	-
0.8479	1800	0.0009	-
0.8714	1850	0.0005	-
0.8950	1900	0.0005	-
0.9185	1950	0.0004	-
0.9421	2000	0.0004	-
0.9656	2050	0.0004	-
0.9892	2100	0.0011	-
1.0	2123	-	0.0231

Framework Versions

• Python: 3.12.12
• SetFit: 1.2.0.dev0
• Sentence Transformers: 5.2.0
• Transformers: 4.57.3
• PyTorch: 2.9.0+cu126
• Datasets: 4.0.0
• Tokenizers: 0.22.2

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}
}