faodl/20250908_model_g20_multilabel_MiniLM-L12-all-labels

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 OneVsRestClassifier 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 OneVsRestClassifier instance
• Maximum Sequence Length: 128 tokens

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	0.3919

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("faodl/20250908_model_g20_multilabel_MiniLM-L12-all-labels")
# Run inference
preds = model("4.3.3 Strategies for Comprehensive Sexuality Education and (CSE) Youth-friendly Health Services 1. To promote volunteerism as a tool for fostering active participation of young people in national development; 5. To promote volunteerism as a tool for fostering active participation of young people in national development; 5.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	70.5122	1194

Training Hyperparameters

• batch_size: (32, 32)
• num_epochs: (2, 2)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 10
• 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
• l2_weight: 0.01
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0005	1	0.1435	-
0.0241	50	0.1438	-
0.0482	100	0.1239	-
0.0723	150	0.1073	-
0.0964	200	0.0992	-
0.1205	250	0.0883	-
0.1446	300	0.08	-
0.1687	350	0.0801	-
0.1928	400	0.073	-
0.2169	450	0.0647	-
0.2410	500	0.0549	-
0.2651	550	0.0575	-
0.2892	600	0.0544	-
0.3133	650	0.0523	-
0.3373	700	0.0506	-
0.3614	750	0.0467	-
0.3855	800	0.0443	-
0.4096	850	0.0385	-
0.4337	900	0.0425	-
0.4578	950	0.0412	-
0.4819	1000	0.036	-
0.5060	1050	0.0323	-
0.5301	1100	0.0352	-
0.5542	1150	0.0347	-
0.5783	1200	0.0319	-
0.6024	1250	0.0254	-
0.6265	1300	0.0291	-
0.6506	1350	0.0253	-
0.6747	1400	0.0283	-
0.6988	1450	0.0248	-
0.7229	1500	0.02	-
0.7470	1550	0.0249	-
0.7711	1600	0.0208	-
0.7952	1650	0.021	-
0.8193	1700	0.0238	-
0.8434	1750	0.0196	-
0.8675	1800	0.0213	-
0.8916	1850	0.0222	-
0.9157	1900	0.019	-
0.9398	1950	0.0226	-
0.9639	2000	0.0156	-
0.9880	2050	0.0193	-
1.0120	2100	0.016	-
1.0361	2150	0.019	-
1.0602	2200	0.0154	-
1.0843	2250	0.0136	-
1.1084	2300	0.014	-
1.1325	2350	0.0147	-
1.1566	2400	0.0126	-
1.1807	2450	0.0161	-
1.2048	2500	0.0123	-
1.2289	2550	0.0151	-
1.2530	2600	0.0123	-
1.2771	2650	0.0122	-
1.3012	2700	0.0084	-
1.3253	2750	0.0154	-
1.3494	2800	0.014	-
1.3735	2850	0.0124	-
1.3976	2900	0.0146	-
1.4217	2950	0.0103	-
1.4458	3000	0.0116	-
1.4699	3050	0.013	-
1.4940	3100	0.0104	-
1.5181	3150	0.0124	-
1.5422	3200	0.0127	-
1.5663	3250	0.0122	-
1.5904	3300	0.0092	-
1.6145	3350	0.0108	-
1.6386	3400	0.0121	-
1.6627	3450	0.0125	-
1.6867	3500	0.0162	-
1.7108	3550	0.0105	-
1.7349	3600	0.0133	-
1.7590	3650	0.0145	-
1.7831	3700	0.0113	-
1.8072	3750	0.009	-
1.8313	3800	0.0105	-
1.8554	3850	0.011	-
1.8795	3900	0.0087	-
1.9036	3950	0.0159	-
1.9277	4000	0.0101	-
1.9518	4050	0.0112	-
1.9759	4100	0.0111	-
2.0	4150	0.0124	-

Framework Versions

• Python: 3.12.11
• SetFit: 1.1.3
• Sentence Transformers: 5.1.0
• Transformers: 4.56.0
• PyTorch: 2.8.0+cu126
• Datasets: 4.0.0
• Tokenizers: 0.22.0

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}
}