anismahmahi/appeal-to-authority-setfit-model

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

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-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/paraphrase-mpnet-base-v2
• 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	'“They know this is one of the great scandals in the history of our country because basically what they did is, they used [former Trump campaign aide] Carter Page, who nobody even knew, who I feel very badly for, I think he’s been treated very badly.\n' 'The Guardian did not make a mistake in vilifying Assange without a shred of evidence.\n' 'He himself said: “No one defends Islam like Arab Christians.” It is to defend Islam that Western clerics do not raise their voice against such acts of brutality.\n'
1	'As the political scientist Richard Neustadt said, political elites are constantly evaluating and re-evaluating the president.\n' '“I can tell you 100% this is not that kind of guy,” said Rick, adding that he would see Paddock every other day and that the two would go to a local bar and play slot machines.\n' 'Now, new information released by investigative reporter Laura Loomer proves that authorities have directly lied to the American people about the case at least once by claiming that supposed shooter Stephen Paddock checked into the Mandalay Bay Hotel on September 28th when valet records (with photos) prove he actually arrived three days earlier.\n'

Evaluation

Metrics

Label	F1
all	0.2237

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("anismahmahi/appeal-to-authority-setfit-model")
# Run inference
preds = model("Ganesh makes wild leaps and inferences.
")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	28.8867	111

Label	Training Sample Count
0	452
1	113

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (2, 2)
• 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
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0007	1	0.3148	-
0.0354	50	0.2792	-
0.0708	100	0.1707	-
0.1062	150	0.1197	-
0.1415	200	0.0768	-
0.1769	250	0.0406	-
0.2123	300	0.0053	-
0.2477	350	0.0571	-
0.2831	400	0.0324	-
0.3185	450	0.001	-
0.3539	500	0.077	-
0.3892	550	0.0002	-
0.4246	600	0.0011	-
0.4600	650	0.003	-
0.4954	700	0.0004	-
0.5308	750	0.0004	-
0.5662	800	0.0006	-
0.6016	850	0.0002	-
0.6369	900	0.0002	-
0.6723	950	0.0003	-
0.7077	1000	0.0116	-
0.7431	1050	0.0059	-
0.7785	1100	0.0002	-
0.8139	1150	0.0001	-
0.8493	1200	0.0001	-
0.8846	1250	0.0003	-
0.9200	1300	0.0001	-
0.9554	1350	0.0	-
0.9908	1400	0.0125	-
1.0	1413	-	0.2868
1.0262	1450	0.0003	-
1.0616	1500	0.0002	-
1.0970	1550	0.0001	-
1.1323	1600	0.0002	-
1.1677	1650	0.0001	-
1.2031	1700	0.0001	-
1.2385	1750	0.0038	-
1.2739	1800	0.0001	-
1.3093	1850	0.0065	-
1.3447	1900	0.0002	-
1.3800	1950	0.0002	-
1.4154	2000	0.0197	-
1.4508	2050	0.0061	-
1.4862	2100	0.0001	-
1.5216	2150	0.0	-
1.5570	2200	0.0321	-
1.5924	2250	0.0002	-
1.6277	2300	0.0331	-
1.6631	2350	0.0069	-
1.6985	2400	0.0001	-
1.7339	2450	0.0	-
1.7693	2500	0.0	-
1.8047	2550	0.0337	-
1.8401	2600	0.0347	-
1.8754	2650	0.0612	-
1.9108	2700	0.0398	-
1.9462	2750	0.0001	-
1.9816	2800	0.0001	-
2.0	2826	-	0.2926

• The bold row denotes the saved checkpoint.

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: 2.16.1
• 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}
}