SetFit with sentence-transformers/all-MiniLM-L6-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/all-MiniLM-L6-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:

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/all-MiniLM-L6-v2
Classification head: a LogisticRegression instance
Maximum Sequence Length: 256 tokens
Number of Classes: 14 classes

Model Sources

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

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("mann2107/BCMPIIRAB_MiniLM_ALLNew")
# Run inference
preds = model("Thank you for your email. Please go ahead and issue. Please invoice in KES")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	25.6577	136

Label	Training Sample Count
0	24
1	24
2	24
3	24
4	24
5	24
6	24
7	24
8	24
9	24
10	24
11	24
12	24
13	24

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (1, 1)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 99
body_learning_rate: (0.0002733656643765287, 0.0002733656643765287)
head_learning_rate: 2.7029049129688732e-05
loss: CosineSimilarityLoss
distance_metric: cosine_distance
margin: 0.25
end_to_end: False
use_amp: False
warmup_proportion: 0.1
max_length: 512
seed: 42
eval_max_steps: -1
load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0002	1	0.2546	-
0.0120	50	0.1667	-
0.0241	100	0.1165	-
0.0361	150	0.0799	-
0.0481	200	0.0212	-
0.0601	250	0.0188	-
0.0722	300	0.0531	-
0.0842	350	0.0273	-
0.0962	400	0.0111	-
0.1082	450	0.0203	-
0.1203	500	0.0397	-
0.1323	550	0.0164	-
0.1443	600	0.0045	-
0.1563	650	0.0032	-
0.1684	700	0.001	-
0.1804	750	0.0011	-
0.1924	800	0.0004	-
0.2044	850	0.0009	-
0.2165	900	0.0006	-
0.2285	950	0.0008	-
0.2405	1000	0.0004	-
0.2525	1050	0.0008	-
0.2646	1100	0.0005	-
0.2766	1150	0.0006	-
0.2886	1200	0.0007	-
0.3006	1250	0.0043	-
0.3127	1300	0.0004	-
0.3247	1350	0.0005	-
0.3367	1400	0.0005	-
0.3487	1450	0.0004	-
0.3608	1500	0.0004	-
0.3728	1550	0.0005	-
0.3848	1600	0.0007	-
0.3968	1650	0.0006	-
0.4089	1700	0.0002	-
0.4209	1750	0.0006	-
0.4329	1800	0.0008	-
0.4449	1850	0.0003	-
0.4570	1900	0.0005	-
0.4690	1950	0.0003	-
0.4810	2000	0.0003	-
0.4930	2050	0.0003	-
0.5051	2100	0.0006	-
0.5171	2150	0.0003	-
0.5291	2200	0.0002	-
0.5411	2250	0.0002	-
0.5532	2300	0.0002	-
0.5652	2350	0.0004	-
0.5772	2400	0.0003	-
0.5892	2450	0.0003	-
0.6013	2500	0.0002	-
0.6133	2550	0.0002	-
0.6253	2600	0.0013	-
0.6373	2650	0.0002	-
0.6494	2700	0.0007	-
0.6614	2750	0.0004	-
0.6734	2800	0.0007	-
0.6854	2850	0.0018	-
0.6975	2900	0.0002	-
0.7095	2950	0.0003	-
0.7215	3000	0.0006	-
0.7335	3050	0.0003	-
0.7456	3100	0.0002	-
0.7576	3150	0.0002	-
0.7696	3200	0.0002	-
0.7816	3250	0.0002	-
0.7937	3300	0.0002	-
0.8057	3350	0.0001	-
0.8177	3400	0.0003	-
0.8297	3450	0.0002	-
0.8418	3500	0.0002	-
0.8538	3550	0.0002	-
0.8658	3600	0.0002	-
0.8778	3650	0.0002	-
0.8899	3700	0.0002	-
0.9019	3750	0.0005	-
0.9139	3800	0.0002	-
0.9259	3850	0.0001	-
0.9380	3900	0.0004	-
0.9500	3950	0.0001	-
0.9620	4000	0.0005	-
0.9740	4050	0.0002	-
0.9861	4100	0.0002	-
0.9981	4150	0.0001	-
1.0	4158	-	0.0302

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
SetFit: 1.1.0.dev0
Sentence Transformers: 3.0.1
Transformers: 4.42.4
PyTorch: 2.3.1+cu121
Datasets: 2.20.0
Tokenizers: 0.19.1

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}
}

mann2107
/

BCMPIIRAB_MiniLM_ALLNew