metadata

library_name: setfit
tags:
  - setfit
  - sentence-transformers
  - text-classification
  - generated_from_setfit_trainer
base_model: intfloat/multilingual-e5-small
metrics:
  - accuracy
widget:
  - text: >-
      query: Sí, la próxima vez que vayas, cuenta conmigo. He querido salir y
      hacer más actividades en la naturaleza.
  - text: 'query: I''m man, I''m leaving now.'
  - text: 'query: Ja, forse possiamo fare un giro in bicicletta insieme.'
  - text: 'query: Mak saya suruh balik, jumpa lagi.'
  - text: 'query: İnanılmaz, bu harika! Bir ayı gördüğüne inanamıyorum!'
pipeline_tag: text-classification
inference: true

SetFit with intfloat/multilingual-e5-small

This is a SetFit model that can be used for Text Classification. This SetFit model uses intfloat/multilingual-e5-small as the Sentence Transformer embedding model. A SetFitHead 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: intfloat/multilingual-e5-small
Classification head: a SetFitHead instance
Maximum Sequence Length: 512 tokens
Number of Classes: 2 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
1	'query: Tja, måste dra nu, ses senare.' 'query: Ispričavam se, moram sada otići.' 'query: Przepraszam, muszę już iść.'
0	'query: Sveiki, kā jums klājas?' 'query: அதிர்ச்சிகரமானது, அது மிகவும் அருமையாக இருக்கிறது! நீ கரடியை பார்த்தது எனக்கு நம்பிக்கையே வரவில்லை!' 'query: Ég hef það fínt, takk. Og þú?'

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("setfit_model_id")
# Run inference
preds = model("query: I'm man, I'm leaving now.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	7.6965	31

Label	Training Sample Count
0	902
1	910

Training Hyperparameters

batch_size: (16, 2)
num_epochs: (1, 16)
max_steps: -1
sampling_strategy: undersampling
body_learning_rate: (1e-05, 1e-05)
head_learning_rate: 0.001
loss: CosineSimilarityLoss
distance_metric: cosine_distance
margin: 0.1
end_to_end: False
use_amp: False
warmup_proportion: 0.1
seed: 42
run_name: multilingual-e5-small
eval_max_steps: -1
load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0000	1	0.3613	-
0.0005	50	0.3577	-
0.0010	100	0.3511	0.3413
0.0015	150	0.3372	-
0.0019	200	0.3447	0.3347
0.0024	250	0.3349	-
0.0029	300	0.3326	0.3224
0.0034	350	0.3372	-
0.0039	400	0.3185	0.3039
0.0044	450	0.2828	-
0.0049	500	0.3055	0.2774
0.0054	550	0.2594	-
0.0058	600	0.2779	0.2489
0.0063	650	0.2486	-
0.0068	700	0.2321	0.22
0.0073	750	0.1838	-
0.0078	800	0.1845	0.2075
0.0083	850	0.1899	-
0.0088	900	0.2147	0.2025
0.0093	950	0.1644	-
0.0097	1000	0.2019	0.1821
0.0102	1050	0.2309	-
0.0107	1100	0.2084	0.1784
0.0112	1150	0.1508	-
0.0117	1200	0.1064	0.1453
0.0122	1250	0.1376	-
0.0127	1300	0.0828	0.121
0.0132	1350	0.1628	-
0.0136	1400	0.1308	0.1018
0.0141	1450	0.0566	-
0.0146	1500	0.0953	0.0767
0.0151	1550	0.1607	-
0.0156	1600	0.1322	0.0625
0.0161	1650	0.0861	-
0.0166	1700	0.0926	0.0423
0.0171	1750	0.0338	-
0.0175	1800	0.1029	0.0344
0.0180	1850	0.0442	-
0.0185	1900	0.019	0.0256
0.0190	1950	0.0489	-
0.0195	2000	0.0675	0.0187

Framework Versions

Python: 3.10.11
SetFit: 1.0.3
Sentence Transformers: 2.7.0
Transformers: 4.39.0
PyTorch: 2.4.0
Datasets: 2.20.0
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}
}