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

This is a SetFit model 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:

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: BAAI/bge-small-en-v1.5
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	'Calcite' 'biotite. Contact metamorphosis' 'rail trail'
1	'Geafitti on tree and burn scar on ground' 'another beautiful rock from the same place' 'Vhfgv'

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("scenic overview")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	7.2788	1899

Label	Training Sample Count
0	2997
1	783

Training Hyperparameters

batch_size: (32, 32)
num_epochs: (1, 1)
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
0.0002	1	0.2331	-
0.0106	50	0.2391	-
0.0212	100	0.238	-
0.0317	150	0.2309	-
0.0423	200	0.2117	-
0.0529	250	0.1879	-
0.0635	300	0.1745	-
0.0741	350	0.1708	-
0.0847	400	0.1402	-
0.0952	450	0.1349	-
0.1058	500	0.1092	-
0.1164	550	0.1031	-
0.1270	600	0.0828	-
0.1376	650	0.0756	-
0.1481	700	0.0587	-
0.1587	750	0.0487	-
0.1693	800	0.0557	-
0.1799	850	0.0456	-
0.1905	900	0.0371	-
0.2011	950	0.0412	-
0.2116	1000	0.0382	-
0.2222	1050	0.0376	-
0.2328	1100	0.0353	-
0.2434	1150	0.0346	-
0.2540	1200	0.0364	-
0.2646	1250	0.0317	-
0.2751	1300	0.0374	-
0.2857	1350	0.0282	-
0.2963	1400	0.0255	-
0.3069	1450	0.023	-
0.3175	1500	0.0287	-
0.3280	1550	0.025	-
0.3386	1600	0.0216	-
0.3492	1650	0.0241	-
0.3598	1700	0.0234	-
0.3704	1750	0.0279	-
0.3810	1800	0.0239	-
0.3915	1850	0.0199	-
0.4021	1900	0.0252	-
0.4127	1950	0.0219	-
0.4233	2000	0.0228	-
0.4339	2050	0.0204	-
0.4444	2100	0.0231	-
0.4550	2150	0.0144	-
0.4656	2200	0.0229	-
0.4762	2250	0.0129	-
0.4868	2300	0.0219	-
0.4974	2350	0.0194	-
0.5079	2400	0.0172	-
0.5185	2450	0.0177	-
0.5291	2500	0.0252	-
0.5397	2550	0.0251	-
0.5503	2600	0.014	-
0.5608	2650	0.0204	-
0.5714	2700	0.0248	-
0.5820	2750	0.0146	-
0.5926	2800	0.0191	-
0.6032	2850	0.0223	-
0.6138	2900	0.0206	-
0.6243	2950	0.0163	-
0.6349	3000	0.0235	-
0.6455	3050	0.0245	-
0.6561	3100	0.0199	-
0.6667	3150	0.0145	-
0.6772	3200	0.016	-
0.6878	3250	0.0143	-
0.6984	3300	0.0206	-
0.7090	3350	0.0187	-
0.7196	3400	0.0168	-
0.7302	3450	0.0148	-
0.7407	3500	0.0212	-
0.7513	3550	0.0185	-
0.7619	3600	0.015	-
0.7725	3650	0.0187	-
0.7831	3700	0.0161	-
0.7937	3750	0.0204	-
0.8042	3800	0.0182	-
0.8148	3850	0.0157	-
0.8254	3900	0.0197	-
0.8360	3950	0.0133	-
0.8466	4000	0.0211	-
0.8571	4050	0.0155	-
0.8677	4100	0.0197	-
0.8783	4150	0.0168	-
0.8889	4200	0.0139	-
0.8995	4250	0.0132	-
0.9101	4300	0.018	-
0.9206	4350	0.014	-
0.9312	4400	0.017	-
0.9418	4450	0.0173	-
0.9524	4500	0.0163	-
0.9630	4550	0.0178	-
0.9735	4600	0.0176	-
0.9841	4650	0.0126	-
0.9947	4700	0.0194	-

Framework Versions

Python: 3.12.9
SetFit: 1.1.2
Sentence Transformers: 4.1.0
Transformers: 4.52.4
PyTorch: 2.7.1
Datasets: 3.6.0
Tokenizers: 0.21.1

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}
}

FriedGil
/

rockd-image-relevance-classification