SetFit with sentence-transformers/all-mpnet-base-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/all-mpnet-base-v2 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: sentence-transformers/all-mpnet-base-v2
Classification head: a LogisticRegression instance
Maximum Sequence Length: 384 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	'To fight bioterrorism sir.' '85V-265V 10W LED Warm White Light Motion Sensor Outdoor Flood Light PIR Lamp AUC http://t.co/NJVPXzMj5V http://t.co/Ijd7WzV5t9' 'Photo: referencereference: xekstrin: I THOUGHT THE NOSTRILS WERE EYES AND I ALMOST CRIED FROM FEAR partake... http://t.co/O7yYjLuKfJ'
1	'Police officer wounded suspect dead after exchanging shots: RICHMOND Va. (AP) \x89ÛÓ A Richmond police officer wa... http://t.co/Y0qQS2L7bS' "There's a weird siren going off here...I hope Hunterston isn't in the process of blowing itself to smithereens..." 'Iranian warship points weapon at American helicopter... http://t.co/cgFZk8Ha1R'

Evaluation

Metrics

Label	Accuracy
all	0.8058

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("pEpOo/catastrophy6")
# Run inference
preds = model("SHOUOUT TO @kasad1lla CAUSE HER VOCALS ARE BLAZING HOT LIKE THE WEATHER SHES IN")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	14.7175	54

Label	Training Sample Count
0	1335
1	948

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (1, 1)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 20
body_learning_rate: (2e-05, 2e-05)
head_learning_rate: 2e-05
loss: CosineSimilarityLoss
distance_metric: cosine_distance
margin: 0.25
end_to_end: False
use_amp: False
warmup_proportion: 0.1
seed: 42
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0094	1	0.0044	-
0.4717	50	0.005	-
0.9434	100	0.0007	-
0.0002	1	0.4675	-
0.0088	50	0.3358	-
0.0175	100	0.2516	-
0.0263	150	0.2158	-
0.0350	200	0.1924	-
0.0438	250	0.1907	-
0.0526	300	0.2166	-
0.0613	350	0.2243	-
0.0701	400	0.0644	-
0.0788	450	0.1924	-
0.0876	500	0.166	-
0.0964	550	0.2117	-
0.1051	600	0.0793	-
0.1139	650	0.0808	-
0.1226	700	0.1183	-
0.1314	750	0.0808	-
0.1402	800	0.0194	-
0.1489	850	0.0699	-
0.1577	900	0.0042	-
0.1664	950	0.0048	-
0.1752	1000	0.1886	-
0.1840	1050	0.0008	-
0.1927	1100	0.0033	-
0.2015	1150	0.0361	-
0.2102	1200	0.12	-
0.2190	1250	0.0035	-
0.2278	1300	0.0002	-
0.2365	1350	0.0479	-
0.2453	1400	0.0568	-
0.2540	1450	0.0004	-
0.2628	1500	0.0002	-
0.2715	1550	0.0013	-
0.2803	1600	0.0005	-
0.2891	1650	0.0014	-
0.2978	1700	0.0004	-
0.3066	1750	0.0008	-
0.3153	1800	0.0616	-
0.3241	1850	0.0003	-
0.3329	1900	0.001	-
0.3416	1950	0.0581	-
0.3504	2000	0.0657	-
0.3591	2050	0.0584	-
0.3679	2100	0.0339	-
0.3767	2150	0.0081	-
0.3854	2200	0.0001	-
0.3942	2250	0.0009	-
0.4029	2300	0.0018	-
0.4117	2350	0.0001	-
0.4205	2400	0.0012	-
0.4292	2450	0.0001	-
0.4380	2500	0.0003	-
0.4467	2550	0.0035	-
0.4555	2600	0.0172	-
0.4643	2650	0.0383	-
0.4730	2700	0.0222	-
0.4818	2750	0.0013	-
0.4905	2800	0.0007	-
0.4993	2850	0.0003	-
0.5081	2900	0.1247	-
0.5168	2950	0.023	-
0.5256	3000	0.0002	-
0.5343	3050	0.0002	-
0.5431	3100	0.0666	-
0.5519	3150	0.0002	-
0.5606	3200	0.0003	-
0.5694	3250	0.0012	-
0.5781	3300	0.0085	-
0.5869	3350	0.0003	-
0.5957	3400	0.0002	-
0.6044	3450	0.0004	-
0.6132	3500	0.013	-
0.6219	3550	0.0089	-
0.6307	3600	0.0001	-
0.6395	3650	0.0002	-
0.6482	3700	0.0039	-
0.6570	3750	0.0031	-
0.6657	3800	0.0009	-
0.6745	3850	0.0002	-
0.6833	3900	0.0002	-
0.6920	3950	0.0001	-
0.7008	4000	0.0	-
0.7095	4050	0.0212	-
0.7183	4100	0.0001	-
0.7270	4150	0.0586	-
0.7358	4200	0.0001	-
0.7446	4250	0.0003	-
0.7533	4300	0.0126	-
0.7621	4350	0.0001	-
0.7708	4400	0.0001	-
0.7796	4450	0.0001	-
0.7884	4500	0.0	-
0.7971	4550	0.0002	-
0.8059	4600	0.0002	-
0.8146	4650	0.0001	-
0.8234	4700	0.0035	-
0.8322	4750	0.0002	-
0.8409	4800	0.0002	-
0.8497	4850	0.0001	-
0.8584	4900	0.0001	-
0.8672	4950	0.0001	-
0.8760	5000	0.0003	-
0.8847	5050	0.0	-
0.8935	5100	0.0041	-
0.9022	5150	0.0001	-
0.9110	5200	0.0001	-
0.9198	5250	0.0001	-
0.9285	5300	0.0001	-
0.9373	5350	0.0001	-
0.9460	5400	0.0001	-
0.9548	5450	0.0001	-
0.9636	5500	0.0001	-
0.9723	5550	0.0001	-
0.9811	5600	0.0002	-
0.9898	5650	0.0271	-
0.9986	5700	0.0	-

Framework Versions

Python: 3.10.12
SetFit: 1.0.1
Sentence Transformers: 2.2.2
Transformers: 4.35.2
PyTorch: 2.1.0+cu121
Datasets: 2.15.0
Tokenizers: 0.15.0

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}
}

pEpOo
/

catastrophy6