SetFit with BAAI/bge-small-en-v1.5

This is a SetFit model that can be used for Text Classification. This SetFit model uses BAAI/bge-small-en-v1.5 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:

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: BAAI/bge-small-en-v1.5
Classification head: a LogisticRegression instance
Maximum Sequence Length: 512 tokens
Number of Classes: 4 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
end_question	'I’m done, we can close the session now.' 'I’ve addressed everything I wanted to, let’s wrap it up.' 'I’ve covered everything I wanted to say and I’m done.'
nothing	'I think Aaron Rodgers is better than Tom Brady' 'I’ve just thought of something else I want to add.' 'There’s another detail I want to bring up.'
wrap_question	'I’ve gone through everything relevant, let me know if there’s more to cover.' 'That’s my take on this, feel free to ask if anything’s unclear.' 'I think I’ve addressed the question in full, but let me know if you need more.'
next_question	'I’ve finished answering this, time for a new topic.' 'I’m ready to explore a different question.' 'I think we’ve exhausted this, what’s next on the agenda?'

Evaluation

Metrics

Label	Accuracy
all	0.9189

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("nksk/Intent_bge-small-en-v1.5_v1.0")
# Run inference
preds = model("Repeat the question for me please")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	8.6319	16

Label	Training Sample Count
end_question	31
next_question	33
nothing	47
wrap_question	33

Training Hyperparameters

batch_size: (32, 16)
num_epochs: (3, 10)
max_steps: -1
sampling_strategy: oversampling
body_learning_rate: (2e-05, 1e-05)
head_learning_rate: 0.0005
loss: CosineSimilarityLoss
distance_metric: cosine_distance
margin: 0.25
end_to_end: True
use_amp: True
warmup_proportion: 0.1
seed: 42
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0021	1	0.2202	-
0.1040	50	0.2429	-
0.2079	100	0.1651	-
0.3119	150	0.0268	-
0.4158	200	0.0079	-
0.5198	250	0.0033	-
0.6237	300	0.0031	-
0.7277	350	0.002	-
0.8316	400	0.0022	-
0.9356	450	0.0022	-
1.0395	500	0.002	-
1.1435	550	0.0017	-
1.2474	600	0.0014	-
1.3514	650	0.001	-
1.4553	700	0.0013	-
1.5593	750	0.0013	-
1.6632	800	0.0011	-
1.7672	850	0.0011	-
1.8711	900	0.0014	-
1.9751	950	0.001	-
2.0790	1000	0.0009	-
2.1830	1050	0.001	-
2.2869	1100	0.0009	-
2.3909	1150	0.0008	-
2.4948	1200	0.0009	-
2.5988	1250	0.0011	-
2.7027	1300	0.0009	-
2.8067	1350	0.0009	-
2.9106	1400	0.0009	-

Framework Versions

Python: 3.10.12
SetFit: 1.0.3
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}
}

nksk
/

Intent_bge-small-en-v1.5_v1.0