brendaac/setfit-prom-bac

SetFit

This is a SetFit model that can be used for Text Classification. 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
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 512 tokens
• Number of Classes: 16 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
Appetite	'I consider myself to be a spiritual person' 'I would feel out of place eating fruits and vegetables at a party where everyone is eating other foods' 'I cook healthy foods'
Future	'I have the impression that the world tends toward collapse (apocalyptic end)' 'I feel anxious about death' 'I am uneasy about possible mishaps'
Concentration	'I have to push myself to get things done because I am tired' 'I have trouble remembering things because I am tired' 'I have enough energy to exercise strenuously'
Financial Worries	'Is your financial condition worsening as a result of your cancer diagnosis or treatment?' 'I am satisfied with my current financial situation' 'I am frustrated that I cannot work or contribute as much as I usually do'
Outcome	'I can make sense of my life' 'I have very clear goals and aims for my life' 'I am confident in this treatment'
Alcohol	'Alcohol makes it easier to talk to people' 'I drank because I had nothing to do' 'Drinking can be harmful to physical health'
Present	'My life is meaningful' 'I feel disconnected from others' 'I make time for family and friends'
Physical Function	'Are you able to climb up five steps?' 'Body pain impedes my exercise participation.' 'Are you able to use a hammer to pound a nail?'
Spiritual Concerns	'I ask people to pray for me' 'My illness has strengthened my faith or spiritual beliefs ' 'My life lacks meaning and purpose '
Drugs	'I did risky things while using drugs' 'When I used drugs, I felt alone' 'I experienced cravings for pain medication'
Loss of Usual Health	'I can rely on my judgment to manage my symptoms, even when others disagree with me' 'I am unable to do the things I want to do because of my medical condition' 'My medical condition makes it hard for me to take care of myself'
Smoking	'I am tempted to smoke when I have arguments or conflicts with others' "When I'm alone, a cigarette can help me pass the time" 'Smoking keeps me from overeating'
Treatment UX	'Overall, how bothered were you by the amount of time it took to take your cancer therapy?' 'I am angry about the lack of care coordination for my healthcare.' "I feel like I don't have access to all the healthcare services I need"
Support	'Do you have someone to help with your daily chores if you are sick?' 'Someone to give you information to help you understand a situation' 'I have someone I trust to talk with about my problems'
Feeling Out of Control	'I have trouble doing all of my usual work (include work at home)' 'How much did pain interfere with your ability to participate in leisure activities?' 'How often was pain distressing to you?'
Sleep	'I felt tired' 'I tossed and turned at night' 'I was satisfied with my sleep'

Evaluation

Metrics

Label	Accuracy
all	0.8715

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("brendaac/setfit-prom-bac")
# Run inference
preds = model("My sleep was deep")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	3	9.625	20

Label	Training Sample Count
Alcohol	8
Appetite	8
Concentration	8
Drugs	8
Feeling Out of Control	8
Financial Worries	8
Future	8
Loss of Usual Health	8
Outcome	8
Physical Function	8
Present	8
Sleep	8
Smoking	8
Spiritual Concerns	8
Support	8
Treatment UX	8

Training Hyperparameters

• batch_size: (32, 32)
• num_epochs: (2, 2)
• max_steps: -1
• sampling_strategy: oversampling
• body_learning_rate: (2e-05, 1e-05)
• head_learning_rate: 0.01
• loss: CosineSimilarityLoss
• distance_metric: cosine_distance
• margin: 0.25
• end_to_end: True
• use_amp: False
• warmup_proportion: 0.1
• l2_weight: 0.01
• seed: 101010
• eval_max_steps: -1
• load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0021	1	0.1955	-
0.1042	50	0.1787	-
0.2083	100	0.1152	-
0.3125	150	0.0475	-
0.4167	200	0.0276	-
0.5208	250	0.0161	-
0.625	300	0.0114	-
0.7292	350	0.0094	-
0.8333	400	0.007	-
0.9375	450	0.0063	-
1.0	480	-	0.0585
1.0417	500	0.0055	-
1.1458	550	0.0051	-
1.25	600	0.0047	-
1.3542	650	0.0043	-
1.4583	700	0.0042	-
1.5625	750	0.004	-
1.6667	800	0.0039	-
1.7708	850	0.0037	-
1.875	900	0.0037	-
1.9792	950	0.0038	-
2.0	960	-	0.0577

Framework Versions

• Python: 3.10.20
• SetFit: 1.1.3
• Sentence Transformers: 3.4.1
• Transformers: 4.41.0
• PyTorch: 2.10.0
• Datasets: 3.3.2
• Tokenizers: 0.19.1

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}
}