Efficient Few-Shot Learning Without Prompts
Paper
•
2209.11055
•
Published
•
4
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:
| Label | Examples |
|---|---|
| 9.0 |
|
| 6.0 |
|
| 5.0 |
|
| 2.0 |
|
| 8.0 |
|
| 3.0 |
|
| 1.0 |
|
| 4.0 |
|
| 0.0 |
|
| 7.0 |
|
| Label | Accuracy |
|---|---|
| all | 1.0 |
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_bc29")
# Run inference
preds = model("첫돌 답례품 스티커 무광사각(21개입)가로3.8x세로6.4cm_백일 출산/육아 > 출산/돌기념품 > 돌잔치용품 > 행사용스티커")
| Training set | Min | Median | Max |
|---|---|---|---|
| Word count | 7 | 15.0529 | 28 |
| Label | Training Sample Count |
|---|---|
| 0.0 | 70 |
| 1.0 | 70 |
| 2.0 | 70 |
| 3.0 | 70 |
| 4.0 | 70 |
| 5.0 | 70 |
| 6.0 | 70 |
| 7.0 | 70 |
| 8.0 | 70 |
| 9.0 | 70 |
| Epoch | Step | Training Loss | Validation Loss |
|---|---|---|---|
| 0.0073 | 1 | 0.4834 | - |
| 0.3650 | 50 | 0.4992 | - |
| 0.7299 | 100 | 0.3772 | - |
| 1.0949 | 150 | 0.1276 | - |
| 1.4599 | 200 | 0.0502 | - |
| 1.8248 | 250 | 0.0345 | - |
| 2.1898 | 300 | 0.022 | - |
| 2.5547 | 350 | 0.0146 | - |
| 2.9197 | 400 | 0.0029 | - |
| 3.2847 | 450 | 0.0002 | - |
| 3.6496 | 500 | 0.0001 | - |
| 4.0146 | 550 | 0.0001 | - |
| 4.3796 | 600 | 0.0001 | - |
| 4.7445 | 650 | 0.0001 | - |
| 5.1095 | 700 | 0.0001 | - |
| 5.4745 | 750 | 0.0001 | - |
| 5.8394 | 800 | 0.0 | - |
| 6.2044 | 850 | 0.0 | - |
| 6.5693 | 900 | 0.0 | - |
| 6.9343 | 950 | 0.0 | - |
| 7.2993 | 1000 | 0.0 | - |
| 7.6642 | 1050 | 0.0 | - |
| 8.0292 | 1100 | 0.0 | - |
| 8.3942 | 1150 | 0.0 | - |
| 8.7591 | 1200 | 0.0 | - |
| 9.1241 | 1250 | 0.0 | - |
| 9.4891 | 1300 | 0.0 | - |
| 9.8540 | 1350 | 0.0 | - |
| 10.2190 | 1400 | 0.0 | - |
| 10.5839 | 1450 | 0.0 | - |
| 10.9489 | 1500 | 0.0 | - |
| 11.3139 | 1550 | 0.0 | - |
| 11.6788 | 1600 | 0.0 | - |
| 12.0438 | 1650 | 0.0 | - |
| 12.4088 | 1700 | 0.0 | - |
| 12.7737 | 1750 | 0.0 | - |
| 13.1387 | 1800 | 0.0 | - |
| 13.5036 | 1850 | 0.0 | - |
| 13.8686 | 1900 | 0.0 | - |
| 14.2336 | 1950 | 0.0 | - |
| 14.5985 | 2000 | 0.0 | - |
| 14.9635 | 2050 | 0.0 | - |
| 15.3285 | 2100 | 0.0 | - |
| 15.6934 | 2150 | 0.0 | - |
| 16.0584 | 2200 | 0.0 | - |
| 16.4234 | 2250 | 0.0 | - |
| 16.7883 | 2300 | 0.0 | - |
| 17.1533 | 2350 | 0.0 | - |
| 17.5182 | 2400 | 0.0 | - |
| 17.8832 | 2450 | 0.0 | - |
| 18.2482 | 2500 | 0.0 | - |
| 18.6131 | 2550 | 0.0 | - |
| 18.9781 | 2600 | 0.0 | - |
| 19.3431 | 2650 | 0.0 | - |
| 19.7080 | 2700 | 0.0 | - |
| 20.0730 | 2750 | 0.0 | - |
| 20.4380 | 2800 | 0.0 | - |
| 20.8029 | 2850 | 0.0 | - |
| 21.1679 | 2900 | 0.0 | - |
| 21.5328 | 2950 | 0.0 | - |
| 21.8978 | 3000 | 0.0 | - |
| 22.2628 | 3050 | 0.0 | - |
| 22.6277 | 3100 | 0.0 | - |
| 22.9927 | 3150 | 0.0 | - |
| 23.3577 | 3200 | 0.0 | - |
| 23.7226 | 3250 | 0.0 | - |
| 24.0876 | 3300 | 0.0 | - |
| 24.4526 | 3350 | 0.0 | - |
| 24.8175 | 3400 | 0.0 | - |
| 25.1825 | 3450 | 0.0 | - |
| 25.5474 | 3500 | 0.0 | - |
| 25.9124 | 3550 | 0.0 | - |
| 26.2774 | 3600 | 0.0 | - |
| 26.6423 | 3650 | 0.0 | - |
| 27.0073 | 3700 | 0.0 | - |
| 27.3723 | 3750 | 0.0 | - |
| 27.7372 | 3800 | 0.0 | - |
| 28.1022 | 3850 | 0.0 | - |
| 28.4672 | 3900 | 0.0 | - |
| 28.8321 | 3950 | 0.0 | - |
| 29.1971 | 4000 | 0.0 | - |
| 29.5620 | 4050 | 0.0 | - |
| 29.9270 | 4100 | 0.0 | - |
@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}
}