priyadp1/setfit-multilingual

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 OneVsRestClassifier 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 OneVsRestClassifier instance
• Maximum Sequence Length: 128 tokens
• Number of Classes: 6 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("setfit_model_id")
# Run inference
preds = model("I just vomited")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	7.9297	17

Training Hyperparameters

• batch_size: (4, 4)
• num_epochs: (2, 2)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 10
• 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	0.3903	-
0.0180	50	0.2544	-
0.0360	100	0.2240	-
0.0541	150	0.2069	-
0.0721	200	0.1946	-
0.0901	250	0.1553	-
0.1081	300	0.1542	-
0.1261	350	0.1490	-
0.1441	400	0.1384	-
0.1622	450	0.1072	-
0.1802	500	0.1045	-
0.1982	550	0.1209	-
0.2162	600	0.0889	-
0.2342	650	0.1010	-
0.2523	700	0.0843	-
0.2703	750	0.0790	-
0.2883	800	0.0734	-
0.3063	850	0.0590	-
0.3243	900	0.0848	-
0.3423	950	0.0520	-
0.3604	1000	0.0523	-
0.3784	1050	0.0680	-
0.3964	1100	0.0716	-
0.4144	1150	0.0732	-
0.4324	1200	0.0584	-
0.4505	1250	0.0600	-
0.4685	1300	0.0522	-
0.4865	1350	0.0489	-
0.5045	1400	0.0524	-
0.5225	1450	0.0524	-
0.5405	1500	0.0507	-
0.5586	1550	0.0666	-
0.5766	1600	0.0575	-
0.5946	1650	0.0490	-
0.6126	1700	0.0487	-
0.6306	1750	0.0518	-
0.6486	1800	0.0554	-
0.6667	1850	0.0394	-
0.6847	1900	0.0517	-
0.7027	1950	0.0494	-
0.7207	2000	0.0413	-
0.7387	2050	0.0652	-
0.7568	2100	0.0427	-
0.7748	2150	0.0354	-
0.7928	2200	0.0586	-
0.8108	2250	0.0452	-
0.8288	2300	0.0465	-
0.8468	2350	0.0422	-
0.8649	2400	0.0349	-
0.8829	2450	0.0391	-
0.9009	2500	0.0404	-
0.9189	2550	0.0382	-
0.9369	2600	0.0376	-
0.9550	2650	0.0530	-
0.9730	2700	0.0426	-
0.9910	2750	0.0377	-
1.0090	2800	0.0419	-
1.0270	2850	0.0399	-
1.0450	2900	0.0374	-
1.0631	2950	0.0245	-
1.0811	3000	0.0348	-
1.0991	3050	0.0332	-
1.1171	3100	0.0310	-
1.1351	3150	0.0315	-
1.1532	3200	0.0355	-
1.1712	3250	0.0329	-
1.1892	3300	0.0312	-
1.2072	3350	0.0400	-
1.2252	3400	0.0370	-
1.2432	3450	0.0287	-
1.2613	3500	0.0349	-
1.2793	3550	0.0390	-
1.2973	3600	0.0275	-
1.3153	3650	0.0388	-
1.3333	3700	0.0299	-
1.3514	3750	0.0286	-
1.3694	3800	0.0293	-
1.3874	3850	0.0361	-
1.4054	3900	0.0326	-
1.4234	3950	0.0278	-
1.4414	4000	0.0317	-
1.4595	4050	0.0330	-
1.4775	4100	0.0252	-
1.4955	4150	0.0377	-
1.5135	4200	0.0223	-
1.5315	4250	0.0352	-
1.5495	4300	0.0317	-
1.5676	4350	0.0344	-
1.5856	4400	0.0282	-
1.6036	4450	0.0320	-
1.6216	4500	0.0302	-
1.6396	4550	0.0336	-
1.6577	4600	0.0288	-
1.6757	4650	0.0358	-
1.6937	4700	0.0303	-
1.7117	4750	0.0365	-
1.7297	4800	0.0253	-
1.7477	4850	0.0334	-
1.7658	4900	0.0347	-
1.7838	4950	0.0301	-
1.8018	5000	0.0314	-
1.8198	5050	0.0325	-
1.8378	5100	0.0304	-
1.8559	5150	0.0290	-
1.8739	5200	0.0269	-
1.8919	5250	0.0312	-
1.9099	5300	0.0262	-
1.9279	5350	0.0323	-
1.9459	5400	0.0301	-
1.9640	5450	0.0232	-
1.9820	5500	0.0304	-
2.0	5550	0.0297	-

Framework Versions

• Python: 3.12.1
• SetFit: 1.2.0
• Sentence Transformers: 6.1.0
• Transformers: 5.19.0
• PyTorch: 2.6.0
• Datasets: 4.4.1
• Tokenizers: 0.23.3

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}
}