SetFit with sentence-transformers/paraphrase-MiniLM-L3-v2
This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-MiniLM-L3-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 |
| 8 |
- 'dev ios available itunes connect dev ready test test open testflight ios device using ios later install using dev agree crash data well statistics use dev provided stash gaming pty ltd linked email stash gaming pty ltd may contact regarding review testflig'
|
| 6 |
- 'updated terms service reaching let know updating figma terms service starter professional figma figma reaching let know updating figma terms service starter professional regularly ensure terms clear cover growing list services available figma also updated '
- 'google play developer program policy update developer update hello google play give users control updating health connect policy strengthen safeguards regarding handling sensitive health record health connect android platform allows health fitness apps sto'
- 'updating privacy notice hello updating privacy notice need believe total process making privacy notice clearer easier changes come effect march find updated privacy account discoverability account discoverability feature allows wise users send money enteri'
|
| 10 |
- 'tumbil fix coderabbitai bot left comment tumbil walkthrough conditional check added analytics event logging methods analyticsservice analytics events logged application running production determined event names parameters remain changes file change summary'
- 'tumbil fix atoruliahotshotslabs approved pull'
- 'tumbil fix comment merge pull request online'
|
| 7 |
- 'review submission review submission eligible submission app maidpro'
|
| 0 |
- 'теніс група sat jun gmt alex sheiko'
|
| 15 |
- 'new arc jobs react native developer ios emea apac found job matches timezone technical skills actively hiring react native developer ios emea apac location timezone open candidates france overlap berlin hourly rate required skills react native ios hours pe'
- 'scalable capital gmbh invites complete codility assessment codility alex reminder technical skills assessment created scalable capital gmbh waiting best please review codility candidate faq also try practice problem familiarize testing please use link begi'
- 'ramanji dasarla sent message days ago ramanji dasarla sent message days ago contacting behalf hirexa reply'
|
| 18 |
- 'dependabot compute migration github actions writing update regarding dependabot hosted compute migration github migration action summary key'
- 'judge rules training copyrighted works fair agentic biology meta befriends alexandr wang united states district court ruled training llms copyrighted books constitutes fair view browser batch top banner june subscribe submit tip mail'
|
| 2 |
- 'zmiana statusu faktury dzień status faktury numerze proforma został zmieniony zaakceptowano jeśli mają państwo pytania lub wątpliwości prosimy kontakt koordynatorem lub pod poznaj nasze fundacja rozwoju przedsiębiorczości twój startup żurawia lok warszawa '
|
| 1 |
- 'easiest way launch scalable web services deploy scalable web services delightful developer experience render makes effortless deploy web apps favorite language python fastapi name every web service render gets add custom deploy web render dashboard click n'
- 'horseworld new account details welcome new details new slack workspace horseworld open slack setting workspace used slack basics ways get even new connect tools sync run share much connecting slack apps slack marketplace notifications keep track projects c'
- 'ready ship'
|
| 14 |
- 'github jfrog unlock unified security powering powers world image alex balancing devops velocity robust security constant join jfrog github deliver unified code binary advanced discover seamlessly embed security across development code without sacrificing r'
- 'weekly post beyond repricing risk real insight weekly post beyond repricing risk real insight weekly post beyond repricing risk real insight making internal pricing ftp competitive advantage market precision accountability internal pricing longer treated t'
- 'bulk download books mathematics education viewed assessing century integrating research findings bulk download books mathematics'
|
| 12 |
- 'build failed encountered error build process means build complete successfully latest changes may exited status optimize view logs learn troubleshooting deploys render team want receive change notification settings workspace need contact support'
|
| 13 |
- 'new device using account new device using account new device signed netflix details device mac chrome web browser location poland location may time june gmt someone enjoy someone please remember allow people household use know recommend change password imm'
- 'security alert google email classifier granted access google account grant check activity secure check activity also see security activity received email let know important changes google account google ireland gordon barrow dublin ireland'
- 'petcube notifications noticed recent login petcube timestamp utc action follow link secure reset password wags petcube team'
|
| 4 |
- 'basecamp uncommon knowledge latest activity work software development project management promotions basecamp uncommon knowledge hours ago hey uncommon knowledge latest activity across everything since june app hypnosis downloads ios android flutter added f'
- 'got unread messages horseworld members team sent messages recently teammates joined view unread messages new messages lisajune need log back lisajune james refresh data brianjune james updating booking top object response bottom would delay cache api repli'
|
| 9 |
- 'join san francisco flutterflow developers conference join san october biggest event year designed enterprise teams building next generation digital deep product insights workshops fireside chats leading voices flutterflow developers conference ffdc chance '
|
| 16 |
- 'vosaio app technology lead writing regarding vosaio wondering might time quick call discuss please let know look forward connecting ali sent outlook mac'
- 'new message waiting flash design teammate trying reach flash design new message vroom karim local seo expert june alex reply instantly mail'
|
| 3 |
- 'payment received payment applied mobilex financial payment information date jun utc amount credited new balance payment method credit card reference primary billing method questions please contact upwork'
|
| 11 |
- 'jamesgilmoursimpson invited work slack jamesgilmoursimpson invited join slack profile picture lisa profile picture profile picture alex profile picture ana profile picture sue profile picture join conversation jamesgilmoursimpson others workspace called jo'
|
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("setfit_model_id")
preds = model("added loading merged")
Training Details
Training Set Metrics
| Training set |
Min |
Median |
Max |
| Word count |
2 |
29.6170 |
42 |
| Label |
Training Sample Count |
| 0 |
1 |
| 1 |
4 |
| 2 |
1 |
| 3 |
1 |
| 4 |
2 |
| 6 |
7 |
| 7 |
1 |
| 8 |
1 |
| 9 |
1 |
| 10 |
3 |
| 11 |
1 |
| 12 |
1 |
| 13 |
3 |
| 14 |
13 |
| 15 |
3 |
| 16 |
2 |
| 18 |
2 |
Training Hyperparameters
- batch_size: (16, 16)
- num_epochs: (1, 16)
- max_steps: 500
- sampling_strategy: oversampling
- body_learning_rate: (1e-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
- max_length: 256
- seed: 42
- eval_max_steps: -1
- load_best_model_at_end: False
Training Results
| Epoch |
Step |
Training Loss |
Validation Loss |
| 0.0164 |
1 |
0.3201 |
- |
| 0.8197 |
50 |
0.1918 |
- |
| 1.6393 |
100 |
0.1084 |
- |
| 0.0141 |
1 |
0.5921 |
- |
| 0.7042 |
50 |
0.4376 |
- |
| 1.4085 |
100 |
0.0449 |
- |
| 2.1127 |
150 |
0.0062 |
- |
| 2.8169 |
200 |
0.0034 |
- |
| 3.5211 |
250 |
0.0024 |
- |
| 4.2254 |
300 |
0.002 |
- |
| 4.9296 |
350 |
0.0016 |
- |
| 5.6338 |
400 |
0.0014 |
- |
| 6.3380 |
450 |
0.0013 |
- |
| 7.0423 |
500 |
0.0013 |
- |
Framework Versions
- Python: 3.13.5
- SetFit: 1.1.2
- Sentence Transformers: 4.1.0
- Transformers: 4.53.0
- PyTorch: 2.7.1
- Datasets: 3.6.0
- Tokenizers: 0.21.2
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}
}
