vidhi0206/setfit-paraphrase-mpnet-sst5_v2

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:

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-mpnet-base-v2
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 512 tokens
• Number of Classes: 5 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	"it 's not a motion picture ; it 's an utterly static picture ." "frankly , it 's kind of insulting , both to men and women ." 'under-rehearsed and lifeless'
2	"recoing 's fantastic performance does n't exactly reveal what makes vincent tick , but perhaps any definitive explanation for it would have felt like a cheat ." "do n't expect any subtlety from this latest entry in the increasingly threadbare gross-out comedy cycle ." "merry friggin ' christmas !"
3	"so purely enjoyable that you might not even notice it 's a fairly straightforward remake of hollywood comedies such as father of the bride ." "what saves this deeply affecting film from being merely a collection of wrenching cases is corcuera 's attention to detail ." 'for once , a movie does not proclaim the truth about two love-struck somebodies , but permits them time and space to convince us of that all on their own .'
1	"the fact that it is n't very good is almost beside the point ." 'what starts off as a satisfying kids flck becomes increasingly implausible as it races through contrived plot points .' 'the film is ultimately about as inspiring as a hallmark card .'
4	'cool gadgets and creatures keep this fresh .' 'morton deserves an oscar nomination .' 'a brutal and funny work .'

Evaluation

Metrics

Label	Accuracy
all	0.5380

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("vidhi0206/setfit-paraphrase-mpnet-sst5_v2")
# Run inference
preds = model("my response to the film is best described as lukewarm .")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	18.8062	52

Label	Training Sample Count
0	64
1	64
2	64
3	64
4	64

Training Hyperparameters

• batch_size: (8, 8)
• num_epochs: (1, 1)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 20
• 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
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0006	1	0.2259	-
0.0312	50	0.2373	-
0.0625	100	0.1726	-
0.0938	150	0.1607	-
0.125	200	0.1869	-
0.1562	250	0.1863	-
0.1875	300	0.224	-
0.2188	350	0.1625	-
0.25	400	0.1284	-
0.2812	450	0.1357	-
0.3125	500	0.2193	-
0.3438	550	0.1434	-
0.375	600	0.0524	-
0.4062	650	0.0558	-
0.4375	700	0.072	-
0.4688	750	0.0312	-
0.5	800	0.0732	-
0.5312	850	0.0117	-
0.5625	900	0.0311	-
0.5938	950	0.0228	-
0.625	1000	0.0026	-
0.6562	1050	0.0196	-
0.6875	1100	0.0017	-
0.7188	1150	0.0067	-
0.75	1200	0.0029	-
0.7812	1250	0.0041	-
0.8125	1300	0.0006	-
0.8438	1350	0.0022	-
0.875	1400	0.0006	-
0.9062	1450	0.0007	-
0.9375	1500	0.001	-
0.9688	1550	0.0009	-
1.0	1600	0.0013	-

Framework Versions

• Python: 3.8.10
• SetFit: 1.0.3
• Sentence Transformers: 2.3.1
• Transformers: 4.37.2
• PyTorch: 2.2.0+cu121
• Datasets: 2.17.0
• Tokenizers: 0.15.1

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}
}