davanstrien/blbooksgenre-setfit

SetFit with BAAI/bge-small-en-v1.5

This is a SetFit model trained on the davanstrien/blbooksgenre dataset that can be used for Text Classification. This SetFit model uses BAAI/bge-small-en-v1.5 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: BAAI/bge-small-en-v1.5
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 512 tokens
• Number of Classes: 2 classes
• Training Dataset: davanstrien/blbooksgenre

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	'Resa i Förenta Staterna' 'The Great Country; or, Impressions of America' "The Visitor's Guide to Oxford. (Abridged ... from the 'Handbook for Visitors to Oxford.') A new edition, with 110 illustrations [The preface is signed, J. P.]"
0	'Cousin Simon [A novel.]' 'Geschied- en oudheidkundige beschrijving van de pleinen, straten, stegen, waterleidingen, wedden, putten en pompen der stad Utrecht, etc' 'The Mystic Quest. A tale of two incarnations'

Evaluation

Metrics

Label	Accuracy
all	0.8793

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("Mirabelle [A novel.]")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	13.0836	77

Label	Training Sample Count
0	348
1	1040

Training Hyperparameters

• batch_size: (32, 32)
• num_epochs: (3, 3)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 20
• 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
0.0006	1	0.2366	-
0.0288	50	0.2442	-
0.0576	100	0.2335	-
0.0865	150	0.2212	-
0.1153	200	0.1858	-
0.1441	250	0.1551	-
0.1729	300	0.1409	-
0.2017	350	0.1148	-
0.2305	400	0.0867	-
0.2594	450	0.0546	-
0.2882	500	0.0326	-
0.3170	550	0.0256	-
0.3458	600	0.0145	-
0.3746	650	0.0137	-
0.4035	700	0.0134	-
0.4323	750	0.0072	-
0.4611	800	0.0144	-
0.4899	850	0.0059	-
0.5187	900	0.0088	-
0.5476	950	0.011	-
0.5764	1000	0.0048	-
0.6052	1050	0.0026	-
0.6340	1100	0.0077	-
0.6628	1150	0.0022	-
0.6916	1200	0.0023	-
0.7205	1250	0.0022	-
0.7493	1300	0.0018	-
0.7781	1350	0.0032	-
0.8069	1400	0.0016	-
0.8357	1450	0.0013	-
0.8646	1500	0.0032	-
0.8934	1550	0.0022	-
0.9222	1600	0.0025	-
0.9510	1650	0.0005	-
0.9798	1700	0.001	-
1.0086	1750	0.0019	-
1.0375	1800	0.0025	-
1.0663	1850	0.0019	-
1.0951	1900	0.0021	-
1.1239	1950	0.0016	-
1.1527	2000	0.0016	-
1.1816	2050	0.0016	-
1.2104	2100	0.0011	-
1.2392	2150	0.0026	-
1.2680	2200	0.0021	-
1.2968	2250	0.0019	-
1.3256	2300	0.001	-
1.3545	2350	0.0015	-
1.3833	2400	0.0028	-
1.4121	2450	0.0023	-
1.4409	2500	0.0016	-
1.4697	2550	0.0024	-
1.4986	2600	0.0035	-
1.5274	2650	0.0021	-
1.5562	2700	0.0015	-
1.5850	2750	0.0009	-
1.6138	2800	0.0009	-
1.6427	2850	0.0018	-
1.6715	2900	0.0004	-
1.7003	2950	0.0006	-
1.7291	3000	0.0004	-
1.7579	3050	0.0011	-
1.7867	3100	0.0004	-
1.8156	3150	0.0004	-
1.8444	3200	0.0006	-
1.8732	3250	0.0005	-
1.9020	3300	0.0003	-
1.9308	3350	0.0004	-
1.9597	3400	0.0003	-
1.9885	3450	0.0003	-
2.0173	3500	0.0003	-
2.0461	3550	0.0003	-
2.0749	3600	0.0003	-
2.1037	3650	0.0003	-
2.1326	3700	0.0003	-
2.1614	3750	0.0003	-
2.1902	3800	0.0003	-
2.2190	3850	0.0009	-
2.2478	3900	0.0003	-
2.2767	3950	0.0003	-
2.3055	4000	0.0003	-
2.3343	4050	0.0003	-
2.3631	4100	0.0003	-
2.3919	4150	0.0003	-
2.4207	4200	0.0003	-
2.4496	4250	0.0003	-
2.4784	4300	0.0003	-
2.5072	4350	0.0003	-
2.5360	4400	0.0003	-
2.5648	4450	0.0003	-
2.5937	4500	0.0003	-
2.6225	4550	0.0003	-
2.6513	4600	0.0003	-
2.6801	4650	0.0003	-
2.7089	4700	0.0003	-
2.7378	4750	0.0003	-
2.7666	4800	0.0003	-
2.7954	4850	0.0003	-
2.8242	4900	0.0003	-
2.8530	4950	0.0003	-
2.8818	5000	0.0003	-
2.9107	5050	0.0003	-
2.9395	5100	0.0003	-
2.9683	5150	0.0003	-
2.9971	5200	0.0003	-

Framework Versions

• Python: 3.12.9
• SetFit: 1.1.3
• Sentence Transformers: 5.3.0
• Transformers: 4.57.6
• PyTorch: 2.11.0
• Datasets: 4.8.4
• Tokenizers: 0.22.2

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}
}