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

This is a SetFit model trained on the rbojja/zero-shot-intent-classification dataset 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: 18 classes
Training Dataset: rbojja/zero-shot-intent-classification

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
7	'Oh my, this is great!' 'Oh, this is fantastic!' 'Hmm, I’m so delighted!'
3	"Oh, absolutely, that's it!" "Oh, absolutely, that's it!" "Yep, that's exactly what I meant."
15	'Really, no way?' 'Oh, that’s quite something!' 'Oh, that’s quite something!'
8	"Gotcha... oh, that's clear!" 'Hmm, I see... perfect!' 'Oh, I see... clear!'
12	'Uhh, fine.' 'Oh, clear.' 'Uhh, noted.'
9	'Uhh, take care!' 'Hmm, see you!' 'Uhh, see you!'
17	'"Umm, this could be a decent plan."' '"I think this might be the solution."' '"Maybe this will work out, I suppose."'
0	"Why can't you just work?!" 'Seriously, this is a joke!' 'Ugh, this is so frustrating!'
6	'"Oh, what if I'm a dream?"' '"Oh, do you speak dolphin?"' '"Uhh, do you have a wish?"'
11	"Uh-huh, that's a valid point." 'Like, I get it.' 'Right, I understand.'
16	'Thank you!' '"Hmmm, thanks, you're great!"' '"Oh, fantastic, thanks a lot!"'
4	"Sorry, I'm not sure." "Well, I'm lost." "Hmm, I'm not sure."
10	'Oh, hi!' "Hello! What's new?" "Hi! How's life?"
13	'Oh, gotcha.' 'Hmmm, okay.' 'Alright, thanks.'
2	'What’s the context behind that?' 'Could you simplify that for me?' 'Can you explain that concept?'
1	'"Oh, I didn’t mean to."' '"Oops, sorry for the oversight."' '"Oops, I’m really sorry."'
5	'Oh, this is not what I wanted.' 'Oh no, this is not right.' 'Seriously, this is a failure.'
14	'Uhh, superb choice!' 'Uhh, amazing decision!' 'Oh, superb performance!'

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("rbojja/intent-classification-small")
# Run inference
preds = model("Uhh, clear.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	4.2224	9

Label	Training Sample Count
0	40
1	40
2	37
3	40
4	41
5	38
6	42
7	38
8	35
9	39
10	42
11	41
12	42
13	44
14	38
15	43
16	47
17	37

Training Hyperparameters

batch_size: (16, 2)
num_epochs: (1, 16)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 20
body_learning_rate: (2e-05, 1e-05)
head_learning_rate: 0.01
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
0.0006	1	0.149	-
0.0276	50	0.1836	-
0.0552	100	0.1408	-
0.0829	150	0.0978	-
0.1105	200	0.0805	-
0.1381	250	0.0684	-
0.1657	300	0.0594	-
0.1934	350	0.051	-
0.2210	400	0.0383	-
0.2486	450	0.0379	-
0.2762	500	0.035	-
0.3039	550	0.0334	-
0.3315	600	0.0306	-
0.3591	650	0.0266	-
0.3867	700	0.0264	-
0.4144	750	0.018	-
0.4420	800	0.0193	-
0.4696	850	0.0166	-
0.4972	900	0.0165	-
0.5249	950	0.016	-
0.5525	1000	0.0177	-
0.5801	1050	0.0202	-
0.6077	1100	0.0133	-
0.6354	1150	0.014	-
0.6630	1200	0.013	-
0.6906	1250	0.0161	-
0.7182	1300	0.0119	-
0.7459	1350	0.0132	-
0.7735	1400	0.0131	-
0.8011	1450	0.0123	-
0.8287	1500	0.0115	-
0.8564	1550	0.0111	-
0.8840	1600	0.011	-
0.9116	1650	0.01	-
0.9392	1700	0.0098	-
0.9669	1750	0.0142	-
0.9945	1800	0.0132	-

Framework Versions

Python: 3.11.11
SetFit: 1.1.1
Sentence Transformers: 3.3.1
Transformers: 4.47.1
PyTorch: 2.5.1+cu121
Datasets: 3.2.0
Tokenizers: 0.21.0

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}
}

rbojja
/

intent-classification-small