reecursion/few-shot-stress-detection-miniLM

SetFit with sentence-transformers/paraphrase-MiniLM-L3-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-MiniLM-L3-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-MiniLM-L3-v2
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 128 tokens
• Number of Classes: 3 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
1	'He is Male, his heart rate is 64, he walks 10000 steps daily, and is Normal. He slept at 11 hrs. Yesterday, he slept from 22.0hrs to 11.0 hrs, with a duration of 765.0 minutes and 2 interruptions. The day before yesterday, he slept from 23.0 hrs to 8.0 hrs, with a duration of 527.0 minutes and 4 interruptions.' 'She is Female, her heart rate is 89, she walks 3873 steps daily and is Overweight. She slept at 10 hrs. Yesterday, she slept from 4.0 hrs to 6.0 hrs, with a duration of 120.0 minutes and 1 interruptions. The day before yesterday, she slept from 4.0 hrs to 9.0 hrs, with a duration of 300.0 minutes and 2 interruptions.' 'She is Female, her heart rate is 68, she walks 11000 steps daily and is Normal. She slept at 10 hrs. Yesterday, she slept from 1.0 hrs to 9.0 hrs, with a duration of 495.0 minutes and 0 interruptions. The day before yesterday, she slept from 1.0 hrs to 10.0 hrs, with a duration of 540.0 minutes and 1 interruptions.'
2	'She is Female, her heart rate is 66, she walks 2413 steps daily and is Underweight. She slept at 8 hrs. Yesterday, she slept from 23.0 hrs to 7.0 hrs, with a duration of 472.0 minutes and 5 interruptions. The day before yesterday, she slept from 23.0 hrs to 5.0 hrs, with a duration of 344.0 minutes and 6 interruptions.' 'He is Male, his heart rate is 95, he walks 9000 steps daily, and is Normal. He slept at 10 hrs. Yesterday, he slept from 4.0hrs to 9.0 hrs, with a duration of 323.0 minutes and 5 interruptions. The day before yesterday, he slept from 2.0 hrs to 10.0 hrs, with a duration of 501.0 minutes and 6 interruptions.'
0	'She is Female, her heart rate is 100, she walks 8000 steps daily and is Normal. She slept at 7 hrs. Yesterday, she slept from 2.0 hrs to 7.0 hrs, with a duration of 323.0 minutes and 0 interruptions. The day before yesterday, she slept from 0.0 hrs to 6.0 hrs, with a duration of 395.0 minutes and 2 interruptions.' 'He is Male, his heart rate is 93, he walks 9800 steps daily, and is Normal. He slept at 9 hrs. Yesterday, he slept from 23.0hrs to 7.0 hrs, with a duration of 460.0 minutes and 0 interruptions. The day before yesterday, he slept from 23.0 hrs to 7.0 hrs, with a duration of 425.0 minutes and 1 interruptions.'

Evaluation

Metrics

Label	Accuracy
all	0.8667

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("reecursion/few-shot-stress-detection-miniLM")
# Run inference
preds = model("He is Male, his heart rate is 75, he walks 11000 steps daily, and is Normal. He slept at 2 hrs. Yesterday, he slept from 3.0hrs to 7.0 hrs, with a duration of 400.0 minutes and 2 interruptions. The day before yesterday, he slept from 1.0 hrs to 8.0 hrs, with a duration of 450.0 minutes and 3 interruptions.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	59	59.5	60

Label	Training Sample Count
0	2
1	6
2	2

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (1, 1)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 15
• 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
0.0526	1	0.3562	-

Framework Versions

• Python: 3.10.12
• SetFit: 1.0.3
• Sentence Transformers: 2.6.1
• Transformers: 4.38.2
• PyTorch: 2.2.1+cu121
• Datasets: 2.18.0
• Tokenizers: 0.15.2

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}
}