pEpOo/catastrophy4

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:

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: 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	'FOOTBALL IS BACK THIS WEEKEND ITS JUST SUNK IN ??????' 'Tried orange aftershock today. My life will never be the same' "Attack on Titan game on PS Vita yay! Can't wait for 2016"
1	'@author_mike Amen today is the Day of Salvation. THX brother Mike for your great encouragement. - http://t.co/cybKsXHF7d Coming US Tsunami' ". @VELDFest announces refunds after Day two's extreme weather evacuation: http://t.co/PP05eTlK7t http://t.co/3Ol8MhhPMa" 'http://t.co/lMA39ZRWoY There is a way which seemeth right unto a man but the end thereof are the ways of death.'

Evaluation

Metrics

Label	Accuracy
all	0.8233

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/catastrophy4")
# Run inference
preds = model("ThisIsFaz: Anti Collision Rear- #technology #cool http://t.co/KEfxTjTAKB Via Techesback #Tech")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	15.0486	30

Label	Training Sample Count
0	836
1	686

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
0.0003	1	0.4126	-
0.0131	50	0.2779	-
0.0263	100	0.2507	-
0.0394	150	0.2475	-
0.0526	200	0.1045	-
0.0657	250	0.2595	-
0.0788	300	0.1541	-
0.0920	350	0.1761	-
0.1051	400	0.0456	-
0.1183	450	0.1091	-
0.1314	500	0.1335	-
0.1445	550	0.0956	-
0.1577	600	0.0583	-
0.1708	650	0.0067	-
0.1840	700	0.0021	-
0.1971	750	0.0057	-
0.2102	800	0.065	-
0.2234	850	0.0224	-
0.2365	900	0.0008	-
0.2497	950	0.1282	-
0.2628	1000	0.1045	-
0.2760	1050	0.001	-
0.2891	1100	0.0005	-
0.3022	1150	0.0013	-
0.3154	1200	0.0007	-
0.3285	1250	0.0015	-
0.3417	1300	0.0007	-
0.3548	1350	0.0027	-
0.3679	1400	0.0006	-
0.3811	1450	0.0001	-
0.3942	1500	0.0009	-
0.4074	1550	0.0002	-
0.4205	1600	0.0004	-
0.4336	1650	0.0003	-
0.4468	1700	0.0013	-
0.4599	1750	0.0004	-
0.4731	1800	0.0007	-
0.4862	1850	0.0001	-
0.4993	1900	0.0001	-
0.5125	1950	0.0476	-
0.5256	2000	0.0561	-
0.5388	2050	0.0009	-
0.5519	2100	0.0381	-
0.5650	2150	0.017	-
0.5782	2200	0.033	-
0.5913	2250	0.0001	-
0.6045	2300	0.0077	-
0.6176	2350	0.0002	-
0.6307	2400	0.0003	-
0.6439	2450	0.0001	-
0.6570	2500	0.0155	-
0.6702	2550	0.0002	-
0.6833	2600	0.0001	-
0.6965	2650	0.031	-
0.7096	2700	0.0215	-
0.7227	2750	0.0002	-
0.7359	2800	0.0002	-
0.7490	2850	0.0001	-
0.7622	2900	0.0001	-
0.7753	2950	0.0001	-
0.7884	3000	0.0001	-
0.8016	3050	0.0001	-
0.8147	3100	0.0001	-
0.8279	3150	0.0001	-
0.8410	3200	0.0001	-
0.8541	3250	0.0001	-
0.8673	3300	0.0001	-
0.8804	3350	0.0001	-
0.8936	3400	0.0	-
0.9067	3450	0.0156	-
0.9198	3500	0.0	-
0.9330	3550	0.0	-
0.9461	3600	0.0001	-
0.9593	3650	0.0208	-
0.9724	3700	0.0	-
0.9855	3750	0.0001	-
0.9987	3800	0.0001	-

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}
}