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 Sources
Model Labels
Label |
Examples |
non_math |
- 'What is the largest ocean on Earth?'
- 'What is the name of the galaxy that contains our solar system?'
- 'What is the name of the ocean on the east coast of the United States?'
|
math |
- 'Which is more: 7 or 9?'
- 'There are 20 chocolates, and you want to share them equally among 4 friends. How many chocolates will each friend get?'
- "If the teacher says 'Alice has 3 more apples than Bob', how can you represent this using numbers and symbols?"
|
Evaluation
Metrics
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
model = SetFitModel.from_pretrained("serdarcaglar/primary-school-math-question")
preds = model("Can you name three different colors?")
Training Details
Training Set Metrics
Training set |
Min |
Median |
Max |
Word count |
1 |
12.4979 |
33 |
Label |
Training Sample Count |
math |
142 |
non_math |
99 |
Training Hyperparameters
- batch_size: (16, 16)
- num_epochs: (1, 1)
- max_steps: -1
- sampling_strategy: oversampling
- num_iterations: 20
- 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
- seed: 42
- eval_max_steps: -1
- load_best_model_at_end: False
Training Results
Epoch |
Step |
Training Loss |
Validation Loss |
0.0017 |
1 |
0.336 |
- |
0.0829 |
50 |
0.1156 |
- |
0.1658 |
100 |
0.0062 |
- |
0.2488 |
150 |
0.0026 |
- |
0.3317 |
200 |
0.0025 |
- |
0.4146 |
250 |
0.0022 |
- |
0.4975 |
300 |
0.0024 |
- |
0.5804 |
350 |
0.0009 |
- |
0.6633 |
400 |
0.0009 |
- |
0.7463 |
450 |
0.0007 |
- |
0.8292 |
500 |
0.0004 |
- |
0.9121 |
550 |
0.0002 |
- |
0.9950 |
600 |
0.0007 |
- |
Framework Versions
- Python: 3.10.12
- SetFit: 1.0.3
- Sentence Transformers: 2.6.1
- Transformers: 4.38.2
- PyTorch: 2.2.1+cu121
- Datasets: 2.18.0
- Tokenizers: 0.15.2
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}
}