metadata
tags:
- setfit
- sentence-transformers
- text-classification
- generated_from_setfit_trainer
widget:
- text: A fantastical portal opening into another dimension, swirling energy.
- text: Analyze the concept of political trust and its importance for governance.
- text: What makes a particular escape room experience engaging and successful?
- text: What is the function of the lymphatic system?
- text: >-
Desenvolva um conto fictício sobre um mapa antigo que guia para um tesouro
cultural perdido.
metrics:
- accuracy
pipeline_tag: text-classification
library_name: setfit
inference: true
base_model: ibm-granite/granite-embedding-107m-multilingual
model-index:
- name: SetFit with ibm-granite/granite-embedding-107m-multilingual
results:
- task:
type: text-classification
name: Text Classification
dataset:
name: Unknown
type: unknown
split: test
metrics:
- type: accuracy
value: 0.8924137931034483
name: Accuracy
As of 28/07/2025, I instead of using this model, a simpler approach would be to just use one of these Gliclass Models, matching the user's prompt against the prompts classes. But this model will remain here nonetheless.
SetFit with ibm-granite/granite-embedding-107m-multilingual
This is a SetFit model that can be used for Text Classification. This SetFit model uses ibm-granite/granite-embedding-107m-multilingual 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: ibm-granite/granite-embedding-107m-multilingual
- Classification head: a LogisticRegression instance
- Maximum Sequence Length: 512 tokens
- Number of Classes: 30 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 |
|---|---|
| sentiment_analysis |
|
| marketing |
|
| entertainment |
|
| image_generation |
|
| complex_reasoning |
|
| education |
|
| mathematics |
|
| biology |
|
| extraction |
|
| engineering |
|
| ethics |
|
| law |
|
| general_knowledge |
|
| geopolitics |
|
| summarization |
|
| healthcare |
|
| spiritual |
|
| coding |
|
| tool |
|
| politics |
|
| business |
|
| creativity |
|
| physics |
|
| psychological |
|
| history |
|
| translation |
|
| basic_reasoning |
|
| finance |
|
| chemistry |
|
| roleplay |
|
Evaluation
Metrics
| Label | Accuracy |
|---|---|
| all | 0.8924 |
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("cnmoro/prompt-router")
# Run inference
preds = model("What is the function of the lymphatic system?")
Training Details
Training Set Metrics
| Training set | Min | Median | Max |
|---|---|---|---|
| Word count | 3 | 11.6859 | 38 |
| Label | Training Sample Count |
|---|---|
| creativity | 176 |
| extraction | 283 |
| image_generation | 173 |
| education | 181 |
| summarization | 174 |
| chemistry | 174 |
| sentiment_analysis | 179 |
| geopolitics | 181 |
| translation | 179 |
| history | 177 |
| coding | 158 |
| politics | 181 |
| healthcare | 178 |
| business | 170 |
| complex_reasoning | 152 |
| psychological | 174 |
| biology | 172 |
| mathematics | 178 |
| marketing | 177 |
| physics | 177 |
| engineering | 176 |
| roleplay | 171 |
| finance | 175 |
| basic_reasoning | 154 |
| ethics | 180 |
| entertainment | 180 |
| tool | 166 |
| law | 173 |
| spiritual | 175 |
| general_knowledge | 170 |
Training Hyperparameters
- batch_size: (8, 8)
- num_epochs: (1, 16)
- max_steps: 2400
- sampling_strategy: oversampling
- 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
- evaluation_strategy: steps
- eval_max_steps: -1
- load_best_model_at_end: True
Training Results
| Epoch | Step | Training Loss | Validation Loss |
|---|---|---|---|
| 0.0004 | 1 | 0.2374 | - |
| 0.0208 | 50 | 0.2111 | - |
| 0.0417 | 100 | 0.2087 | - |
| 0.0625 | 150 | 0.1995 | - |
| 0.0833 | 200 | 0.1984 | 0.1876 |
| 0.1042 | 250 | 0.1894 | - |
| 0.125 | 300 | 0.1872 | - |
| 0.1458 | 350 | 0.1818 | - |
| 0.1667 | 400 | 0.1758 | 0.1587 |
| 0.1875 | 450 | 0.1647 | - |
| 0.2083 | 500 | 0.1547 | - |
| 0.2292 | 550 | 0.1404 | - |
| 0.25 | 600 | 0.1342 | 0.1252 |
| 0.2708 | 650 | 0.1309 | - |
| 0.2917 | 700 | 0.1209 | - |
| 0.3125 | 750 | 0.1329 | - |
| 0.3333 | 800 | 0.1068 | 0.1055 |
| 0.3542 | 850 | 0.1131 | - |
| 0.375 | 900 | 0.1006 | - |
| 0.3958 | 950 | 0.1033 | - |
| 0.4167 | 1000 | 0.1005 | 0.0922 |
| 0.4375 | 1050 | 0.1133 | - |
| 0.4583 | 1100 | 0.0898 | - |
| 0.4792 | 1150 | 0.0918 | - |
| 0.5 | 1200 | 0.0983 | 0.0855 |
| 0.5208 | 1250 | 0.0947 | - |
| 0.5417 | 1300 | 0.0921 | - |
| 0.5625 | 1350 | 0.1045 | - |
| 0.5833 | 1400 | 0.09 | 0.0763 |
| 0.6042 | 1450 | 0.0893 | - |
| 0.625 | 1500 | 0.0823 | - |
| 0.6458 | 1550 | 0.0853 | - |
| 0.6667 | 1600 | 0.0881 | 0.0713 |
| 0.6875 | 1650 | 0.0837 | - |
| 0.7083 | 1700 | 0.0886 | - |
| 0.7292 | 1750 | 0.0784 | - |
| 0.75 | 1800 | 0.0838 | 0.0680 |
| 0.7708 | 1850 | 0.0743 | - |
| 0.7917 | 1900 | 0.0788 | - |
| 0.8125 | 1950 | 0.084 | - |
| 0.8333 | 2000 | 0.0772 | 0.0659 |
| 0.8542 | 2050 | 0.0872 | - |
| 0.875 | 2100 | 0.0808 | - |
| 0.8958 | 2150 | 0.0649 | - |
| 0.9167 | 2200 | 0.0795 | 0.0651 |
| 0.9375 | 2250 | 0.0774 | - |
| 0.9583 | 2300 | 0.0687 | - |
| 0.9792 | 2350 | 0.0787 | - |
| 1.0 | 2400 | 0.0786 | 0.0647 |
Framework Versions
- Python: 3.11.11
- SetFit: 1.2.0.dev0
- Sentence Transformers: 5.0.0
- Transformers: 4.53.2
- PyTorch: 2.7.1+cu126
- Datasets: 3.2.0
- Tokenizers: 0.21.0
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}
}