ehsanhallo/setfit-paraphrase-multilingual-MiniLM-L12-v2-ig-fa

SetFit with sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-multilingual-MiniLM-L12-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:

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: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 128 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	'چه سودایی که سر همینا از دست دادم😂' 'خو فارسی بنویس بفهمه 😂😂😂😂😂' 'اینجا ایران همین سایتا هم\u200cزیادی..نیازی به بررسی ندارن...کلا دوسداریم به همچی ایراد بگیریم.'
1	'کد کارت مشکی NHKDKI' 'اتفاقا مسیولیت بیشتری برات میاره و درگیریات بیشتر میشه برای هدفی که داری' 'من میخام شروع کنم،اورج بفروشم یا فیک؟فیک ارزونتره ولی فیکه.اورجینال هم ک گرون تره ؟بنظرت اورج میخرن؟؟'
2	'🔥🔥🔥🔥' '😂😂😂' 'چه قدر عالی وخفن 🔥🔥'

Evaluation

Metrics

Label	Accuracy
all	0.79

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("ehsanhallo/setfit-paraphrase-multilingual-MiniLM-L12-v2-ig-fa")
# Run inference
preds = model(" where`d you go!")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	6.4184	75

Label	Training Sample Count
0	69
1	238
2	551

Training Hyperparameters

• batch_size: (32, 16)
• num_epochs: (1, 2)
• max_steps: -1
• sampling_strategy: oversampling
• body_learning_rate: (2e-05, 5e-06)
• head_learning_rate: 0.002
• 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: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0001	1	0.1767	-
0.0216	250	0.1513	-
0.0431	500	0.0629	0.2389
0.0647	750	0.0351	-
0.0862	1000	0.0015	0.1886
0.1078	1250	0.0003	-
0.1293	1500	0.0004	0.1813
0.1509	1750	0.0002	-
0.1724	2000	0.0002	0.1807
0.1940	2250	0.0001	-
0.2155	2500	0.0001	0.187
0.2371	2750	0.0001	-
0.2586	3000	0.0001	0.1903
0.2802	3250	0.0001	-
0.3018	3500	0.0	0.1864
0.3233	3750	0.0	-
0.3449	4000	0.0	0.193
0.3664	4250	0.0	-
0.3880	4500	0.0	0.1879
0.4095	4750	0.0	-
0.4311	5000	0.0	0.1887
0.4526	5250	0.0	-
0.4742	5500	0.0	0.187
0.4957	5750	0.0	-
0.5173	6000	0.0001	0.205
0.5388	6250	0.0	-
0.5604	6500	0.0	0.205
0.5819	6750	0.0	-
0.6035	7000	0.0	0.2018
0.6251	7250	0.0	-
0.6466	7500	0.0	0.2022
0.6682	7750	0.0	-
0.6897	8000	0.0	0.2063
0.7113	8250	0.0	-
0.7328	8500	0.0	0.2143
0.7544	8750	0.0	-
0.7759	9000	0.0	0.2206
0.7975	9250	0.0	-
0.8190	9500	0.0	0.2167
0.8406	9750	0.0	-
0.8621	10000	0.0	0.2176
0.8837	10250	0.0	-
0.9053	10500	0.0	0.217
0.9268	10750	0.0	-
0.9484	11000	0.0	0.2153
0.9699	11250	0.0	-
0.9915	11500	0.0	0.2137

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• SetFit: 1.0.1
• Sentence Transformers: 2.2.2
• Transformers: 4.35.2
• PyTorch: 2.1.0+cu121
• Datasets: 2.16.1
• Tokenizers: 0.15.0

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}
}