vinulacs/sinmix-setfit-sentiment

SetFit with intfloat/multilingual-e5-large

This is a SetFit model that can be used for Text Classification. This SetFit model uses intfloat/multilingual-e5-large 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:

1. Fine-tuning a Sentence Transformer with contrastive learning.
2. Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

• Model Type: SetFit
• Sentence Transformer body: intfloat/multilingual-e5-large
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 512 tokens
• Number of Classes: 3 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
0	'Fiber nae.' 'hey Zemlak boru bill danne ai ewa refund karana kiwata reply na anthima horu tikakne' 'hey Zemlak boru bill danne ai ewa refund karana kiwata reply na horu tikakne'
2	'I think this is the first time Pubg players are awarded with this much of a cash pool in SL.' 'I think this is the first time players are with this of a pool in' 'I think this is the first time Pubg players are awarded with in this much of a cash pool in SL.'
1	'Koo ampara district alamkulam area 2g coverage' 'Koo ampara district area 2g coverage' 'Koo ampara district district alamkulam area 2g coverage'

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("vinulacs/sinmix-setfit-sentiment")
# Run inference
preds = model("Gampola Lebsack chater")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	9.7390	79

Label	Training Sample Count
0	194
1	185
2	165

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (1, 1)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 40
• 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
0.0004	1	0.2675	-
0.0184	50	0.3054	-
0.0368	100	0.219	-
0.0551	150	0.1664	-
0.0735	200	0.0785	-
0.0919	250	0.0256	-
0.1103	300	0.009	-
0.1287	350	0.0104	-
0.1471	400	0.0004	-
0.1654	450	0.0003	-
0.1838	500	0.0003	-
0.2022	550	0.0002	-
0.2206	600	0.0002	-
0.2390	650	0.0002	-
0.2574	700	0.0002	-
0.2757	750	0.0002	-
0.2941	800	0.0001	-
0.3125	850	0.0001	-
0.3309	900	0.0001	-
0.3493	950	0.0001	-
0.3676	1000	0.0001	-
0.3860	1050	0.0001	-
0.4044	1100	0.0001	-
0.4228	1150	0.0001	-
0.4412	1200	0.0001	-
0.4596	1250	0.0001	-
0.4779	1300	0.0001	-
0.4963	1350	0.0001	-
0.5147	1400	0.0001	-
0.5331	1450	0.0001	-
0.5515	1500	0.0001	-
0.5699	1550	0.0001	-
0.5882	1600	0.0001	-
0.6066	1650	0.0001	-
0.625	1700	0.0001	-
0.6434	1750	0.0001	-
0.6618	1800	0.0001	-
0.6801	1850	0.0001	-
0.6985	1900	0.0001	-
0.7169	1950	0.0001	-
0.7353	2000	0.0001	-
0.7537	2050	0.0001	-
0.7721	2100	0.0001	-
0.7904	2150	0.0001	-
0.8088	2200	0.0001	-
0.8272	2250	0.0001	-
0.8456	2300	0.0001	-
0.8640	2350	0.0001	-
0.8824	2400	0.0001	-
0.9007	2450	0.0001	-
0.9191	2500	0.0001	-
0.9375	2550	0.0013	-
0.9559	2600	0.0001	-
0.9743	2650	0.0001	-
0.9926	2700	0.0001	-

Framework Versions

• Python: 3.12.12
• SetFit: 1.1.3
• Sentence Transformers: 5.2.3
• Transformers: 4.51.3
• PyTorch: 2.10.0+cu128
• Datasets: 4.0.0
• Tokenizers: 0.21.4

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}
}