SetFit with sentence-transformers/paraphrase-mpnet-base-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-mpnet-base-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/paraphrase-mpnet-base-v2
Classification head: a LogisticRegression instance
Maximum Sequence Length: 512 tokens
Number of Classes: 2 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
1	"I'd say I eat about half of what I used to. Small, frequent meals work better for me. [SEP] that's good to hear. How often does this loss of appetite happen ?" 'Maybe a couple times a day. No specific times. [SEP] Have you noticed any impact on your work or academic performance?' "I wake up early, as always. Haven't needed more sleep lately. [SEP] that's good. it's important to get enough rest when you're injured. Are there specific factors or worries waking you up early?"
0	"Thanks, I will. I'm going to keep an eye on my weight and appetite in the meantime. [SEP] Have you been experiencing any nausea along with your anxiety symptoms?" 'Fine, nothing really to complain about today. [SEP] How is your digestion going? Do you have any problems?' "I'm really not comfortable discussing this with you anymore. This feels impersonal and dismissive. [SEP] How is your bowel regularity? Do you have a bowel movement every day?"

Label

Examples

"I'd say I eat about half of what I used to. Small, frequent meals work better for me. [SEP] that's good to hear. How often does this loss of appetite happen ?"
'Maybe a couple times a day. No specific times. [SEP] Have you noticed any impact on your work or academic performance?'
"I wake up early, as always. Haven't needed more sleep lately. [SEP] that's good. it's important to get enough rest when you're injured. Are there specific factors or worries waking you up early?"

"Thanks, I will. I'm going to keep an eye on my weight and appetite in the meantime. [SEP] Have you been experiencing any nausea along with your anxiety symptoms?"
'Fine, nothing really to complain about today. [SEP] How is your digestion going? Do you have any problems?'
"I'm really not comfortable discussing this with you anymore. This feels impersonal and dismissive. [SEP] How is your bowel regularity? Do you have a bowel movement every day?"

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("AliaeAI/setfit_nli_v2")
# Run inference
preds = model("I stressed a lot [SEP] What kind of work do you do?")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	6	33.4038	95

Label	Training Sample Count
0	1856
1	1856

Training Hyperparameters

batch_size: (32, 32)
num_epochs: (3, 3)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 10
body_learning_rate: (2e-05, 2e-05)
head_learning_rate: 2e-05
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	0.2774	-
0.0216	50	0.2696	-
0.0431	100	0.2591	-
0.0647	150	0.2611	-
0.0862	200	0.2555	-
0.1078	250	0.2551	-
0.1293	300	0.2579	-
0.1509	350	0.2557	-
0.1724	400	0.2503	-
0.1940	450	0.2553	-
0.2155	500	0.2508	-
0.2371	550	0.2463	-
0.2586	600	0.2405	-
0.2802	650	0.2268	-
0.3017	700	0.2245	-
0.3233	750	0.2057	-
0.3448	800	0.2019	-
0.3664	850	0.2028	-
0.3879	900	0.1716	-
0.4095	950	0.1675	-
0.4310	1000	0.1463	-
0.4526	1050	0.1417	-
0.4741	1100	0.1259	-
0.4957	1150	0.1102	-
0.5172	1200	0.1008	-
0.5388	1250	0.0958	-
0.5603	1300	0.0947	-
0.5819	1350	0.0906	-
0.6034	1400	0.0785	-
0.625	1450	0.0757	-
0.6466	1500	0.0654	-
0.6681	1550	0.0588	-
0.6897	1600	0.0666	-
0.7112	1650	0.0536	-
0.7328	1700	0.0587	-
0.7543	1750	0.0552	-
0.7759	1800	0.0475	-
0.7974	1850	0.0406	-
0.8190	1900	0.0386	-
0.8405	1950	0.0334	-
0.8621	2000	0.0362	-
0.8836	2050	0.0279	-
0.9052	2100	0.0271	-
0.9267	2150	0.0325	-
0.9483	2200	0.0281	-
0.9698	2250	0.0365	-
0.9914	2300	0.0316	-
1.0129	2350	0.024	-
1.0345	2400	0.0237	-
1.0560	2450	0.0244	-
1.0776	2500	0.0217	-
1.0991	2550	0.0183	-
1.1207	2600	0.0175	-
1.1422	2650	0.0169	-
1.1638	2700	0.0233	-
1.1853	2750	0.019	-
1.2069	2800	0.023	-
1.2284	2850	0.0177	-
1.25	2900	0.0158	-
1.2716	2950	0.0195	-
1.2931	3000	0.0098	-
1.3147	3050	0.0202	-
1.3362	3100	0.0094	-
1.3578	3150	0.0178	-
1.3793	3200	0.0168	-
1.4009	3250	0.0184	-
1.4224	3300	0.0132	-
1.4440	3350	0.0139	-
1.4655	3400	0.0132	-
1.4871	3450	0.0131	-
1.5086	3500	0.0147	-
1.5302	3550	0.012	-
1.5517	3600	0.0134	-
1.5733	3650	0.011	-
1.5948	3700	0.0141	-
1.6164	3750	0.0078	-
1.6379	3800	0.0115	-
1.6595	3850	0.0123	-
1.6810	3900	0.0119	-
1.7026	3950	0.0143	-
1.7241	4000	0.0112	-
1.7457	4050	0.01	-
1.7672	4100	0.0139	-
1.7888	4150	0.0113	-
1.8103	4200	0.0093	-
1.8319	4250	0.0091	-
1.8534	4300	0.0124	-
1.875	4350	0.0085	-
1.8966	4400	0.009	-
1.9181	4450	0.0103	-
1.9397	4500	0.008	-
1.9612	4550	0.008	-
1.9828	4600	0.0108	-
2.0043	4650	0.0096	-
2.0259	4700	0.0086	-
2.0474	4750	0.0062	-
2.0690	4800	0.0048	-
2.0905	4850	0.006	-
2.1121	4900	0.0052	-
2.1336	4950	0.0062	-
2.1552	5000	0.0076	-
2.1767	5050	0.0084	-
2.1983	5100	0.0051	-
2.2198	5150	0.0063	-
2.2414	5200	0.0067	-
2.2629	5250	0.0058	-
2.2845	5300	0.0058	-
2.3060	5350	0.0079	-
2.3276	5400	0.0076	-
2.3491	5450	0.0101	-
2.3707	5500	0.0044	-
2.3922	5550	0.0051	-
2.4138	5600	0.0044	-
2.4353	5650	0.0043	-
2.4569	5700	0.0066	-
2.4784	5750	0.0059	-
2.5	5800	0.0097	-
2.5216	5850	0.0054	-
2.5431	5900	0.0057	-
2.5647	5950	0.0033	-
2.5862	6000	0.0049	-
2.6078	6050	0.0038	-
2.6293	6100	0.0056	-
2.6509	6150	0.006	-
2.6724	6200	0.0061	-
2.6940	6250	0.0031	-
2.7155	6300	0.0059	-
2.7371	6350	0.004	-
2.7586	6400	0.0033	-
2.7802	6450	0.0031	-
2.8017	6500	0.0062	-
2.8233	6550	0.0063	-
2.8448	6600	0.0055	-
2.8664	6650	0.0026	-
2.8879	6700	0.004	-
2.9095	6750	0.0039	-
2.9310	6800	0.005	-
2.9526	6850	0.0064	-
2.9741	6900	0.0058	-
2.9957	6950	0.0057	-

Framework Versions

Python: 3.11.13
SetFit: 1.2.0.dev0
Sentence Transformers: 4.1.0
Transformers: 4.52.4
PyTorch: 2.6.0+cu124
Datasets: 3.6.0
Tokenizers: 0.21.1

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}
}

AliaeAI
/

setfit_nli_v2