richie-ghost/setfit-mental-bert-base-uncased-MH-Topic-Check

SetFit with mental/mental-bert-base-uncased

This is a SetFit model that can be used for Text Classification. This SetFit model uses mental/mental-bert-base-uncased 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: mental/mental-bert-base-uncased
• 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
True	'I have so many issues to address. I have a history of sexual abuse, I’m a breast cancer survivor and I am a lifetime insomniac. I have a long history of depression and I’m beginning to have anxiety. I have low self esteem but I’ve been happily married for almost 35 years.\n I’ve never had counseling about any of this. Do I have too many issues to address in counseling?' 'I have so many issues to address. I have a history of sexual abuse, I’m a breast cancer survivor and I am a lifetime insomniac. I have a long history of depression and I’m beginning to have anxiety. I have low self esteem but I’ve been happily married for almost 35 years.\n I’ve never had counseling about any of this. Do I have too many issues to address in counseling?' 'Experiencing extreme mood swings not related to external circumstances.'
False	'Guide to learning a new language' 'Learning about the historical significance of the Silk Road.' 'Exploring historical landmarks in Europe'

Evaluation

Metrics

Label	Accuracy
all	0.9882

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("richie-ghost/setfit-mental-bert-base-uncased-MH-Topic-Check")
# Run inference
preds = model("Planning a DIY home renovation project.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	4	33.7092	111

Label	Training Sample Count
True	138
False	58

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (3, 3)
• max_steps: -1
• sampling_strategy: oversampling
• 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
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0007	1	0.2132	-
0.0354	50	0.1508	-
0.0708	100	0.0193	-
0.1062	150	0.0075	-
0.1415	200	0.0025	-
0.1769	250	0.0009	-
0.2123	300	0.0003	-
0.2477	350	0.0005	-
0.2831	400	0.0004	-
0.3185	450	0.0004	-
0.3539	500	0.0002	-
0.3892	550	0.0004	-
0.4246	600	0.0001	-
0.4600	650	0.0003	-
0.4954	700	0.0001	-
0.5308	750	0.0001	-
0.5662	800	0.0001	-
0.6016	850	0.0002	-
0.6369	900	0.0001	-
0.6723	950	0.0001	-
0.7077	1000	0.0001	-
0.7431	1050	0.0	-
0.7785	1100	0.0001	-
0.8139	1150	0.0001	-
0.8493	1200	0.0001	-
0.8846	1250	0.0001	-
0.9200	1300	0.0001	-
0.9554	1350	0.0001	-
0.9908	1400	0.0001	-
1.0	1413	-	0.017
1.0262	1450	0.0001	-
1.0616	1500	0.0001	-
1.0970	1550	0.0	-
1.1323	1600	0.0001	-
1.1677	1650	0.0001	-
1.2031	1700	0.0001	-
1.2385	1750	0.0	-
1.2739	1800	0.0001	-
1.3093	1850	0.0	-
1.3447	1900	0.0	-
1.3800	1950	0.0	-
1.4154	2000	0.0	-
1.4508	2050	0.0	-
1.4862	2100	0.0	-
1.5216	2150	0.0	-
1.5570	2200	0.0	-
1.5924	2250	0.0	-
1.6277	2300	0.0	-
1.6631	2350	0.0	-
1.6985	2400	0.0	-
1.7339	2450	0.0	-
1.7693	2500	0.0	-
1.8047	2550	0.0	-
1.8401	2600	0.0	-
1.8754	2650	0.0	-
1.9108	2700	0.0001	-
1.9462	2750	0.0	-
1.9816	2800	0.0	-
2.0	2826	-	0.018
2.0170	2850	0.0	-
2.0524	2900	0.0	-
2.0878	2950	0.0	-
2.1231	3000	0.0	-
2.1585	3050	0.0	-
2.1939	3100	0.0	-
2.2293	3150	0.0	-
2.2647	3200	0.0	-
2.3001	3250	0.0	-
2.3355	3300	0.0	-
2.3708	3350	0.0	-
2.4062	3400	0.0	-
2.4416	3450	0.0	-
2.4770	3500	0.0	-
2.5124	3550	0.0	-
2.5478	3600	0.0	-
2.5832	3650	0.0	-
2.6185	3700	0.0	-
2.6539	3750	0.0	-
2.6893	3800	0.0	-
2.7247	3850	0.0	-
2.7601	3900	0.0	-
2.7955	3950	0.0	-
2.8309	4000	0.0	-
2.8662	4050	0.0001	-
2.9016	4100	0.0	-
2.9370	4150	0.0	-
2.9724	4200	0.0001	-
3.0	4239	-	0.0182

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• SetFit: 1.0.3
• Sentence Transformers: 2.7.0
• Transformers: 4.40.0
• PyTorch: 2.2.1+cu121
• Datasets: 2.19.0
• 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}
}