SetFit with cambridgeltl/SapBERT-from-PubMedBERT-fulltext

This is a SetFit model that can be used for Text Classification. This SetFit model uses cambridgeltl/SapBERT-from-PubMedBERT-fulltext as the Sentence Transformer embedding model. A MultiOutputClassifier 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: cambridgeltl/SapBERT-from-PubMedBERT-fulltext
Classification head: a MultiOutputClassifier instance
Maximum Sequence Length: 512 tokens

Model Sources

Repository: SetFit on GitHub
Paper: Efficient Few-Shot Learning Without Prompts
Blogpost: SetFit: Efficient Few-Shot Learning Without Prompts

Evaluation

Metrics

Label	Accuracy
all	0.1757

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("setfit_model_id")
# Run inference
preds = model("caspases")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	1.7652	5

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (3, 3)
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
l2_weight: 0.01
seed: 42
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0009	1	0.2361	-
0.0463	50	0.2377	-
0.0927	100	0.2269	-
0.1390	150	0.2104	-
0.1854	200	0.1871	-
0.2317	250	0.1437	-
0.2780	300	0.1322	-
0.3244	350	0.1365	-
0.3707	400	0.1155	-
0.4171	450	0.1144	-
0.4634	500	0.1068	-
0.5097	550	0.1011	-
0.5561	600	0.095	-
0.6024	650	0.0933	-
0.6487	700	0.1063	-
0.6951	750	0.0999	-
0.7414	800	0.0823	-
0.7878	850	0.0877	-
0.8341	900	0.0767	-
0.8804	950	0.0849	-
0.9268	1000	0.0796	-
0.9731	1050	0.0877	-
1.0195	1100	0.0759	-
1.0658	1150	0.0705	-
1.1121	1200	0.0728	-
1.1585	1250	0.0738	-
1.2048	1300	0.0767	-
1.2512	1350	0.0692	-
1.2975	1400	0.0697	-
1.3438	1450	0.0639	-
1.3902	1500	0.0729	-
1.4365	1550	0.0759	-
1.4829	1600	0.0786	-
1.5292	1650	0.0618	-
1.5755	1700	0.0722	-
1.6219	1750	0.0719	-
1.6682	1800	0.072	-
1.7146	1850	0.0654	-
1.7609	1900	0.0683	-
1.8072	1950	0.0654	-
1.8536	2000	0.0679	-
1.8999	2050	0.0643	-
1.9462	2100	0.0662	-
1.9926	2150	0.0642	-
2.0389	2200	0.0812	-
2.0853	2250	0.068	-
2.1316	2300	0.0583	-
2.1779	2350	0.0627	-
2.2243	2400	0.0654	-
2.2706	2450	0.0571	-
2.3170	2500	0.0623	-
2.3633	2550	0.0639	-
2.4096	2600	0.059	-
2.4560	2650	0.0637	-
2.5023	2700	0.0675	-
2.5487	2750	0.0696	-
2.5950	2800	0.0669	-
2.6413	2850	0.0633	-
2.6877	2900	0.0606	-
2.7340	2950	0.0609	-
2.7804	3000	0.054	-
2.8267	3050	0.0598	-
2.8730	3100	0.0597	-
2.9194	3150	0.0618	-
2.9657	3200	0.065	-

Framework Versions

Python: 3.10.12
SetFit: 1.1.0
Sentence Transformers: 3.1.1
Transformers: 4.39.0
PyTorch: 2.4.1+cu121
Datasets: 3.0.0
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}
}

MiguelRod
/

SetFit-Chemical-Biomaterials-Classifier