SetFit with sentence-transformers/all-MiniLM-L6-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/all-MiniLM-L6-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:

Fine-tuning a Sentence Transformer with contrastive learning.
Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

Model Type: SetFit
Sentence Transformer body: sentence-transformers/all-MiniLM-L6-v2
Classification head: a LogisticRegression instance
Maximum Sequence Length: 256 tokens
Number of Classes: 11 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
Other Factors	'Due to lack of education of the parents of the children, they are unable to send their children to school.' "In some families, girls' education is not prioritized. Due to social stereotypes, household pressures, and a lack of awareness, girls' school attendance and enrollment are relatively low. As a result, girls are being deprived of education at the elementary level, which is a serious concern for their future." 'Due to development, the surrounding environment is not good and there is no awareness about education.'
Early Marriage	'Child marriage is another challenge facing this community.' 'Girls end up marrying their own people, which stops their education and limits their future prospects.' 'Child marriage'
Unknown/Unclear	'Children are enrolled in school but do not attend school regularly.' 'The importance of education in life and the progress of children in their country and their own lives if they are educated were discussed.' 'Being sick disrupts education'
Distance and Accessibility Issues	'Children are unable to attend school because the school is located on the edge of the forest.' 'The school is far away from the village.' 'The school is located in a secluded area, making it difficult for children to reach there.'
Teacher Capacity and Quality Issues	'There is no study in school' 'The rural woman said that there are English teachers in the schools, but the children do not know English.' 'In school, teachers make the child do cleaning work and since the school is very far away, the date of birth is not available.'
Poverty and Economic Barriers	'Children work in the fields with their parents.' 'Child Labor' 'Instead of going to school, children are being sent to graze goats.'
Safety Concerns	'Girls are unable to attend school regularly due to harassment on the way.' 'The head master discriminates against girls from lower caste by saying that these people are from lower caste, he gives them some favour in food and drinks and there is a lot of beating.' 'Parents, distressed by the harsh treatment meted out to girls, are afraid to send their children to school.'
Legal Document linked Barriers	'Children are not getting enrolled in school due to lack of Aadhaar card.' 'Some girls in her village go to school, some have left after enrolling, some do not go to school due to lack of Aadhar card.' 'Admission in school is not possible due to lack of Aadhar card'
Parental Attitudes and Socio-Cultural Barriers	'Lack of coordination between girls and parents' 'Girls take wrong steps and this is also the reason why higher education is wasted.' 'Children do not go to school because of household chores.'
Substance Abuse and Addiction	'Parents challenged that their children do not go to school and are always busy playing games and using mobile phones.' "Children's studies are being disrupted due to misuse of mobile phones." 'Fathers who drink alcohol often neglect their children.'
School Infrastructure and Facility Issues	'In Shiksha Chaupal, the parents were told that there is only a school up to class 5 in our village.' 'Lack of proper infrastructure for education.' 'When we go to Anganwadi, the Didi does not make us take care of food and cleanliness.'

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("setfit_model_id")
# Run inference
preds = model("Early marriage of girls")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	15.3202	158

Label	Training Sample Count
Distance and Accessibility Issues	500
Early Marriage	500
Legal Document linked Barriers	500
Other Factors	500
Parental Attitudes and Socio-Cultural Barriers	500
Poverty and Economic Barriers	500
Safety Concerns	500
School Infrastructure and Facility Issues	500
Substance Abuse and Addiction	500
Teacher Capacity and Quality Issues	500
Unknown/Unclear	500

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (1, 1)
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	0.2641	-
0.0727	50	0.2312	-
0.1453	100	0.2069	-
0.2180	150	0.1816	-
0.2907	200	0.155	-
0.3634	250	0.1408	-
0.4360	300	0.1306	-
0.5087	350	0.1371	-
0.5814	400	0.1301	-
0.6541	450	0.1239	-
0.7267	500	0.1236	-
0.7994	550	0.1213	-
0.8721	600	0.122	-
0.9448	650	0.1298	-

Framework Versions

Python: 3.12.3
SetFit: 1.1.3
Sentence Transformers: 5.6.0
Transformers: 4.57.6
PyTorch: 2.13.0+cpu
Datasets: 2.16.1
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}
}