Efficient Few-Shot Learning Without Prompts
Paper
•
2209.11055
•
Published
•
4
SetFit with mini1013/master_domain
This is a SetFit model that can be used for Text Classification. This SetFit model uses mini1013/master_domain 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: mini1013/master_domain
- Classification head: a LogisticRegression instance
- Maximum Sequence Length: 512 tokens
- Number of Classes: 18 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 |
|---|---|
| 17.0 |
|
| 7.0 |
|
| 13.0 |
|
| 0.0 |
|
| 12.0 |
|
| 16.0 |
|
| 14.0 |
|
| 2.0 |
|
| 3.0 |
|
| 6.0 |
|
| 4.0 |
|
| 9.0 |
|
| 5.0 |
|
| 11.0 |
|
| 8.0 |
|
| 15.0 |
|
| 1.0 |
|
| 10.0 |
|
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("mini1013/master_cate_bc13")
# Run inference
preds = model("꼬마실로폰 완구 원목블록 출산 매트 육아 출산/육아 > 완구/인형 > 음악/악기놀이")
Training Details
Training Set Metrics
| Training set | Min | Median | Max |
|---|---|---|---|
| Word count | 7 | 14.3937 | 25 |
| Label | Training Sample Count |
|---|---|
| 0.0 | 20 |
| 1.0 | 20 |
| 2.0 | 20 |
| 3.0 | 20 |
| 4.0 | 20 |
| 5.0 | 20 |
| 6.0 | 8 |
| 7.0 | 20 |
| 8.0 | 20 |
| 9.0 | 20 |
| 10.0 | 20 |
| 11.0 | 20 |
| 12.0 | 20 |
| 13.0 | 20 |
| 14.0 | 20 |
| 15.0 | 20 |
| 16.0 | 20 |
| 17.0 | 20 |
Training Hyperparameters
- batch_size: (256, 256)
- num_epochs: (30, 30)
- max_steps: -1
- sampling_strategy: oversampling
- num_iterations: 50
- 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 |
|---|---|---|---|
| 0.0147 | 1 | 0.4769 | - |
| 0.7353 | 50 | 0.4673 | - |
| 1.4706 | 100 | 0.1769 | - |
| 2.2059 | 150 | 0.0467 | - |
| 2.9412 | 200 | 0.012 | - |
| 3.6765 | 250 | 0.0068 | - |
| 4.4118 | 300 | 0.0067 | - |
| 5.1471 | 350 | 0.0043 | - |
| 5.8824 | 400 | 0.0022 | - |
| 6.6176 | 450 | 0.0001 | - |
| 7.3529 | 500 | 0.0001 | - |
| 8.0882 | 550 | 0.0001 | - |
| 8.8235 | 600 | 0.0001 | - |
| 9.5588 | 650 | 0.0001 | - |
| 10.2941 | 700 | 0.0001 | - |
| 11.0294 | 750 | 0.0001 | - |
| 11.7647 | 800 | 0.0 | - |
| 12.5 | 850 | 0.0 | - |
| 13.2353 | 900 | 0.0 | - |
| 13.9706 | 950 | 0.0 | - |
| 14.7059 | 1000 | 0.0 | - |
| 15.4412 | 1050 | 0.0 | - |
| 16.1765 | 1100 | 0.0 | - |
| 16.9118 | 1150 | 0.0 | - |
| 17.6471 | 1200 | 0.0 | - |
| 18.3824 | 1250 | 0.0 | - |
| 19.1176 | 1300 | 0.0 | - |
| 19.8529 | 1350 | 0.0 | - |
| 20.5882 | 1400 | 0.0 | - |
| 21.3235 | 1450 | 0.0 | - |
| 22.0588 | 1500 | 0.0 | - |
| 22.7941 | 1550 | 0.0 | - |
| 23.5294 | 1600 | 0.0 | - |
| 24.2647 | 1650 | 0.0 | - |
| 25.0 | 1700 | 0.0 | - |
| 25.7353 | 1750 | 0.0 | - |
| 26.4706 | 1800 | 0.0 | - |
| 27.2059 | 1850 | 0.0 | - |
| 27.9412 | 1900 | 0.0 | - |
| 28.6765 | 1950 | 0.0 | - |
| 29.4118 | 2000 | 0.0 | - |
Framework Versions
- Python: 3.10.12
- SetFit: 1.1.0
- Sentence Transformers: 3.3.1
- Transformers: 4.44.2
- PyTorch: 2.2.0a0+81ea7a4
- Datasets: 3.2.0
- 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}
}
- Downloads last month
- -