lucienbaumgartner/mentalizing-class

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

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/all-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:

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/all-mpnet-base-v2
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 384 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
1	'i panicked and made chatgpt write everything . ' 'it is fundamental that chatgpt with developer mode can say anything about anyone , at any time for any reason . ' "chatgpt itself mentioned that homebrewing requires good system mastery and an understanding of the developer 's game philosophy to make properly balanced feats . "
0	'chatgpt confirmed it . ' "chatgpt does n't know that it writing text that is easily detected . " "the timing of entering the initial prompt is essential to ensure that chatgpt understands the user 's request and can provide an accurate response . "
2	'3 . diversion : chatgpt might also create a diversion , directing a group of wasps to move away from the nest and act as a decoy . ' 'chatgpt can generate content on a wide range of subjects , so the possibilities are endless . ' 'does anyone know if chatgpt can generate the code of a sound wave , with the specifications that are requested , as it does with programming codes . '

Evaluation

Metrics

Label	Accuracy	Precision	Recall	F1
all	0.75	0.7667	0.7460	0.7488

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("chatgpt makes choices , algorithms are n't neutral . ")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	3	20.7848	51

Label	Training Sample Count
0	26
1	27
2	26

Training Hyperparameters

• batch_size: (32, 2)
• num_epochs: (10, 10)
• 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
• l2_weight: 0.01
• seed: 42
• evaluation_strategy: epoch
• eval_max_steps: -1
• load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0077	1	0.2555	-
0.3846	50	0.2528	-
0.7692	100	0.1993	-
1.0	130	-	0.1527
1.1538	150	0.0222	-
1.5385	200	0.0023	-
1.9231	250	0.0013	-
2.0	260	-	0.1461
2.3077	300	0.0015	-
2.6923	350	0.0005	-
3.0	390	-	0.1465
3.0769	400	0.0003	-
3.4615	450	0.0002	-
3.8462	500	0.0003	-
4.0	520	-	0.1353
4.2308	550	0.0007	-
4.6154	600	0.0002	-
5.0	650	0.0011	0.1491
5.3846	700	0.0002	-
5.7692	750	0.0002	-
6.0	780	-	0.1478
6.1538	800	0.0002	-
6.5385	850	0.0001	-
6.9231	900	0.0001	-
7.0	910	-	0.1472
7.3077	950	0.0001	-
7.6923	1000	0.0001	-
8.0	1040	-	0.1461
8.0769	1050	0.0001	-
8.4615	1100	0.0001	-
8.8462	1150	0.0001	-
9.0	1170	-	0.1393
9.2308	1200	0.0001	-
9.6154	1250	0.0001	-
10.0	1300	0.0001	0.1399

Framework Versions

• Python: 3.11.7
• SetFit: 1.1.0
• Sentence Transformers: 3.2.0
• Transformers: 4.45.2
• PyTorch: 2.4.1
• Datasets: 3.0.1
• Tokenizers: 0.20.1

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}
}