Finnish-actions/SetFit-FinBERT1-A2-appreciation

SetFit with TurkuNLP/bert-base-finnish-cased-v1

This is a SetFit model that can be used for Text Classification. This SetFit model uses TurkuNLP/bert-base-finnish-cased-v1 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:

1. Fine-tuning a Sentence Transformer with contrastive learning.
2. Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

• Model Type: SetFit
• Sentence Transformer body: TurkuNLP/bert-base-finnish-cased-v1
• Classification head: a LogisticRegression 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
0	'Etunimi Sukunimi herra senkun aloittaa keräyksen♥️' 'Etunimi Sukunimi venäjän syy hintojen nousu vai syytätkö sodastakin Suomen hallitusta ? 😖' 'Etunimi Sukunimi Olikhaan se virve'
1	'Etunimi Sukunimi onneks sentään ryyppäämään pääsee, eikä tule siihen ikäviä taukoja' 'Etunimi Sukunimi Juurikin näin 👍👍👍👍' 'Etunimi Oinonen-Matikkala, sama tunne jäi. Viisaita, pohdittuja sanoja.'

Evaluation

Metrics

Label	Metric
all	0.9622

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("Finnish-actions/SetFit-FinBERT1-A2-appreciation")
# Run inference
preds = model("Etunimi Sukunimi 🙋‍♀️")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	20.3115	213

Label	Training Sample Count
0	934
1	29

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (4, 4)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 6
• 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: epoch
• eval_max_steps: -1
• load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0014	1	0.2634	-
0.0692	50	0.2672	-
0.1383	100	0.2311	-
0.2075	150	0.1393	-
0.2766	200	0.0183	-
0.3458	250	0.0122	-
0.4149	300	0.0096	-
0.4841	350	0.0136	-
0.5533	400	0.0047	-
0.6224	450	0.0075	-
0.6916	500	0.0041	-
0.7607	550	0.0048	-
0.8299	600	0.0045	-
0.8990	650	0.0015	-
0.9682	700	0.0033	-
1.0	723	-	0.3577
1.0373	750	0.002	-
1.1065	800	0.0055	-
1.1757	850	0.0052	-
1.2448	900	0.0032	-
1.3140	950	0.004	-
1.3831	1000	0.0051	-
1.4523	1050	0.0067	-
1.5214	1100	0.0036	-
1.5906	1150	0.0013	-
1.6598	1200	0.0047	-
1.7289	1250	0.0062	-
1.7981	1300	0.0022	-
1.8672	1350	0.0066	-
1.9364	1400	0.0044	-
2.0	1446	-	0.3291
2.0055	1450	0.0037	-
2.0747	1500	0.0037	-
2.1438	1550	0.003	-
2.2130	1600	0.0004	-
2.2822	1650	0.0042	-
2.3513	1700	0.0016	-
2.4205	1750	0.0033	-
2.4896	1800	0.0034	-
2.5588	1850	0.004	-
2.6279	1900	0.0023	-
2.6971	1950	0.006	-
2.7663	2000	0.0014	-
2.8354	2050	0.0039	-
2.9046	2100	0.0051	-
2.9737	2150	0.0038	-
3.0	2169	-	0.3276
3.0429	2200	0.0024	-
3.1120	2250	0.0022	-
3.1812	2300	0.0022	-
3.2503	2350	0.0032	-
3.3195	2400	0.003	-
3.3887	2450	0.0024	-
3.4578	2500	0.0039	-
3.5270	2550	0.0048	-
3.5961	2600	0.0048	-
3.6653	2650	0.0045	-
3.7344	2700	0.0039	-
3.8036	2750	0.0022	-
3.8728	2800	0.0031	-
3.9419	2850	0.0019	-
4.0	2892	-	0.3279

Framework Versions

• Python: 3.11.9
• SetFit: 1.1.3
• Sentence Transformers: 3.2.0
• Transformers: 4.44.0
• PyTorch: 2.4.0+cu124
• Datasets: 2.21.0
• Tokenizers: 0.19.1

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}
}