This is a SetFit model that can be used for Text Classification. This SetFit model uses google/embeddinggemma-300m 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: google/embeddinggemma-300m
Classification head: a LogisticRegression instance
Maximum Sequence Length: 2048 tokens
Number of Classes: 4 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
neutral	'in this context, canadian authorities are co-operating closely to monitor the financial situation of the household sector, and are responding appropriately.' 'in contrast to the united states, canada’s economy continues to operate with material excess capacity.' 'the july projection also incorporates the estimated impact of tariffs on steel and aluminum recently imposed by the united states, as well as the countermeasures enacted by canada.'
dovish	'in the face of strong global demand for goods, pandemic-related disruptions to production and transportation are constraining growth.' 'business investment in the energy-producing sector will decline.' 'as higher interest rates continue to work their way through the economy, the bank expects economic growth to slow, averaging around 1% through the second half of this year and the first half of next year.'
hawkish	'over the next few months, inflation is expected to rise temporarily to around the top of the 1-3 percent inflation-control range.' 'the three main upside risks relate to the possibility of stronger-than-expected inflationary pressures in the global economy, stronger momentum in canadian household spending, and the possibility of a faster-than-expected rebound in business and consumer confidence, due to more decisive policy action in the major advanced economies.' 'moreover, with three-month rates of core inflation running around 3½-4% since last september, underlying price pressures appear to be more persistent than anticipated.'
irrelevant	'the situation calls for special actions by the central bank.' 'canada has not been immune to these developments.' 'both of these programs will be put in place in the coming weeks.'

Evaluation

Metrics

Label	Accuracy
all	0.9892

Uses

Direct Use for Inference

First install the SetFit library:

pip install setfit

Then you can load this model and run inference.

python

from setfit import SetFitModel

# Download from the 🤗 Hub
model = SetFitModel.from_pretrained("beethogedeon/stance-boc-google-embed-300m")
# Run inference
preds = model("these factors are weighing on growth in many emerging markets and some other economies.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	4	21.0398	57

Label	Training Sample Count
dovish	288
hawkish	249
irrelevant	12
neutral	432

Training Hyperparameters

batch_size: (16, 2)
num_epochs: (1, 16)
max_steps: 500
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
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.002	1	0.2742	-
0.1	50	0.2427	-
0.2	100	0.2168	-
0.3	150	0.1799	-
0.4	200	0.1462	-
0.5	250	0.1066	-
0.6	300	0.0723	-
0.7	350	0.0435	-
0.8	400	0.029	-
0.9	450	0.0196	-
1.0	500	0.0101	-

Framework Versions

Python: 3.12.12
SetFit: 1.1.3
Sentence Transformers: 5.2.0
Transformers: 4.57.3
PyTorch: 2.9.0+cu126
Datasets: 4.0.0
Tokenizers: 0.22.1

Citation

BibTeX

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}
}