tmp-org/tmp_cv_model_2025_10_06_0

SetFit with Alibaba-NLP/gte-multilingual-base

This is a SetFit model trained on the tmp-org/edeka-dataset-ctx-1 dataset that can be used for Text Classification. This SetFit model uses Alibaba-NLP/gte-multilingual-base 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: Alibaba-NLP/gte-multilingual-base
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 8192 tokens
• Number of Classes: 23 classes
• Training Dataset: tmp-org/edeka-dataset-ctx-1

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
Other_Prospekt	'Prospekt () [TextView
Start_Start	'Tierischer Sammelspaß mit schleich® und WWF! () [TextView
Other_Code einlösen	'Hallo Murmuras! () [TextView
Prämien_Prämien	'Noch 5 Treuepunkte zum Prämien-Sonderpreis. () [TextView
Other_Loading	'Noch 5 Treuepunkte zum Prämien-Sonderpreis. () [TextView
Other_Treueaktionen	'Kaufe 4 Packungen 3/4er & Mini - bekomme Coupon für 1 Gratis-Artikel () [TextView
Other_Neuigkeiten	'EDEKA Schöck () [TextView
Other_Produktherkunft	'Aktuelles () [TextView
Other_Marktsuche	'Produktherkunft () [TextView
Other_Menu	'PAYBACK () [TextView
Kasse_Mobil bezahlen	' (Tabelle schließen) [View
Other_Kassenbons	' (Tabelle schließen) [View
Sparen_Angebote	" (Aktivierte Coupons) [View
Sparen_Coupons	'PAYBACK () [TextView
Other_Coupon details	' (coupon image) [ImageView
Kasse_Kasse	'Coupons () [TextView
Kasse_Aktivierte Coupons	' (Aktivierte Coupons) [View
Einkaufsliste_Einkaufsliste	'Coupons () [TextView
Other_Unknown	'Prospekt () [TextView
Start_Loading	'Gratisartikel im November: Gut&Günstig – Classic Spülmittel Apfelduft () [TextView
Sparen_Loading	'Zukünftige Treueaktionen () [TextView
Other_Other	'Obst & Gemüse () [TextView
Kasse_Loading	'Mobil bezahlen per App () [TextView

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("tmp-org/tmp_cv_model_2025_10_06_0")
# Run inference
preds = model("Prospekt () [TextView|SubActivity] | 1 / 12 () [TextView|SubActivity]

[SELECTED START]

Prospekt () [TextView|SubActivity] | 1 / 12 () [TextView|SubActivity]
[CONTEXT SEPARATOR]
Prospekt () [TextView|SubActivity] | 1 / 12 () [TextView|SubActivity]")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	9	273.4604	709

Label	Training Sample Count
Einkaufsliste_Einkaufsliste	31
Kasse_Aktivierte Coupons	40
Kasse_Kasse	20
Kasse_Loading	2
Kasse_Mobil bezahlen	6
Other_Code einlösen	2
Other_Coupon details	32
Other_Kassenbons	8
Other_Loading	1
Other_Marktsuche	6
Other_Menu	40
Other_Neuigkeiten	8
Other_Other	1
Other_Produktherkunft	12
Other_Prospekt	33
Other_Treueaktionen	36
Other_Unknown	10
Prämien_Prämien	38
Sparen_Angebote	40
Sparen_Coupons	40
Sparen_Loading	3
Start_Loading	5
Start_Start	40

Training Hyperparameters

• batch_size: (4, 4)
• num_epochs: (1, 1)
• max_steps: -1
• sampling_strategy: undersampling
• 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: 4242
• eval_max_steps: -1
• load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0003	1	0.1538	-
0.0132	50	0.2586	-
0.0264	100	0.1891	-
0.0396	150	0.1552	-
0.0528	200	0.1651	-
0.0660	250	0.1258	-
0.0792	300	0.1179	-
0.0924	350	0.0903	-
0.1056	400	0.0894	-
0.1188	450	0.1039	-
0.1320	500	0.0899	-
0.1452	550	0.0983	-
0.1584	600	0.0573	-
0.1715	650	0.0735	-
0.1847	700	0.0643	-
0.1979	750	0.0616	-
0.2111	800	0.0754	-
0.2243	850	0.0516	-
0.2375	900	0.062	-
0.2507	950	0.0675	-
0.2639	1000	0.0657	-
0.2771	1050	0.0471	-
0.2903	1100	0.0428	-
0.3035	1150	0.0266	-
0.3167	1200	0.0306	-
0.3299	1250	0.0446	-
0.3431	1300	0.0429	-
0.3563	1350	0.0359	-
0.3695	1400	0.0391	-
0.3827	1450	0.0569	-
0.3959	1500	0.0446	-
0.4091	1550	0.0311	-
0.4223	1600	0.0383	-
0.4355	1650	0.0358	-
0.4487	1700	0.0454	-
0.4619	1750	0.0319	-
0.4751	1800	0.0481	-
0.4883	1850	0.0459	-
0.5015	1900	0.047	-
0.5146	1950	0.0348	-
0.5278	2000	0.0352	-
0.5410	2050	0.0294	-
0.5542	2100	0.0385	-
0.5674	2150	0.0343	-
0.5806	2200	0.0369	-
0.5938	2250	0.035	-
0.6070	2300	0.0188	-
0.6202	2350	0.0301	-
0.6334	2400	0.0438	-
0.6466	2450	0.0295	-
0.6598	2500	0.0292	-
0.6730	2550	0.0208	-
0.6862	2600	0.0195	-
0.6994	2650	0.0263	-
0.7126	2700	0.0391	-
0.7258	2750	0.0252	-
0.7390	2800	0.032	-
0.7522	2850	0.0194	-
0.7654	2900	0.0331	-
0.7786	2950	0.0171	-
0.7918	3000	0.0307	-
0.8050	3050	0.0236	-
0.8182	3100	0.0361	-
0.8314	3150	0.0096	-
0.8446	3200	0.0265	-
0.8577	3250	0.0251	-
0.8709	3300	0.0384	-
0.8841	3350	0.0196	-
0.8973	3400	0.0157	-
0.9105	3450	0.0271	-
0.9237	3500	0.0206	-
0.9369	3550	0.0128	-
0.9501	3600	0.016	-
0.9633	3650	0.0115	-
0.9765	3700	0.0162	-
0.9897	3750	0.0175	-

Framework Versions

• Python: 3.12.6
• SetFit: 1.1.2
• Sentence Transformers: 5.2.2
• Transformers: 4.57.1
• PyTorch: 2.10.0+cu128
• Datasets: 3.6.0
• 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}
}