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 SetFitHead 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/paraphrase-mpnet-base-v2
Classification head: a SetFitHead instance
Maximum Sequence Length: 512 tokens
Number of Classes: 9 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
6	'The man claps his hands together. The man' 'Emerging in open water, he does a breaststroke toward the murky. He' 'The girl does 2 perfect flips. The girls'
3	'The younger insurance rep solemnly faces his partner. The older man' 'He grabs her hair and pulls her head back. She' 'A kid in blue shorts is vacuuming the floor. A kid in a red shirt'
2	'In slow motion, both the Russians and Americans celebrate. Someone' 'Through a window, we watch someone raise his teacup to his companions. At home, someone' 'As our view retracts through the star map a holographic line sets out from the gunner chair and targets hologram of the planet earth. She'
4	"The waiter refills someone's glass. Someone" "He finds someone's records in a box. Someone" "Bloodstains spread over someone's white shirt. Someone"
7	'Now, someone stands below an overcast sky. Strands of his greasy black hair' 'Someone turns at the sound of the distant horns. 6000 horsemen, lead by people,' 'Someone points his wand upwards. High above, red sparks'
5	'Now in the eating quarters, someone faces a husky, larged - nosed cook. The cook' 'A logo for a sports even is shown. There' 'Someone stirs the cookie dough in a bowl. The dough'
0	'He shows a water bottle he has along with a brush, and uses the brush to remove snow from the dash window of a car and the water to remove any excess snow left on the windshield. Once finished, he' "Someone and someone step into a tent. Someone's mouth" 'Someone steps outside and opens an umbrella. Someone halts,'
8	'Someone peers out from the cabin. As she emerges, someone' 'He gently tries to pull up and then reel the fishing line out of the hole. He' 'A woman smiles at the camera. The woman'
1	'We see a title screen. We' 'A lot of people are sitting on terraces in a big field and people is walking in the entrance of a big stadium. men' 'We see the finished painting and a line of paints. We then'

Evaluation

Metrics

Label	Accuracy
all	0.1656

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("HelgeKn/Swag-multi-class-8")
# Run inference
preds = model("He approaches the object and reads a plaque on its side. Someone")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	6	14.0833	40

Label	Training Sample Count
0	8
1	8
2	8
3	8
4	8
5	8
6	8
7	8
8	8

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (2, 2)
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
0.0056	1	0.2013	-
0.2778	50	0.1955	-
0.5556	100	0.0693	-
0.8333	150	0.0166	-
1.1111	200	0.0369	-
1.3889	250	0.0149	-
1.6667	300	0.0095	-
1.9444	350	0.0238	-

Framework Versions

Python: 3.9.13
SetFit: 1.0.1
Sentence Transformers: 2.2.2
Transformers: 4.36.0
PyTorch: 2.1.1+cpu
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}
}

HelgeKn
/

Swag-multi-class-8