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 Sources

Model Labels

Label Examples
Out of Scope
  • 'Why is your website so slow?'
  • 'Can I get a shoutout on your social media?'
  • 'I like to listen to classical music'
product faq
  • 'What is the price of the Temple Butidaar Multi Color Border Pure Silk Chiffon Georgette Saree?'
  • 'Do you have the Air Jordan 1 Low Shadow Brown/Brown Kelp- Sail in size 7?'
  • 'Is the lakadong turmeric powder available for purchase?'
order tracking
  • 'What is the expected delivery time for the 10 pack of Cake Boxes to Bhopal?'
  • 'What is the delivery status for my order placed using email address [email protected]?'
  • 'I havent received my order'
product policy
  • 'What is the policy for returning a product that was part of a Cyber Monday sale?'
  • 'Are there any exceptions to the return policy for items that were purchased with a special occasion promotion?'
  • 'Are there any restrictions on returning sneakers with added fur or fur trim?'
product discoverability
  • 'Suggest me some high ankle sneakers'
  • 'Do you have any grocery & gourmet honey available?'
  • 'Do you have any sneaker collaborations with artists?'
general faq
  • 'How many cups of green tea should I drink daily to achieve the recommended therapeutic dosage of ECGC?'
  • 'what is mashru silk'
  • 'What specific compounds in Green Tea contribute to its antioxidant properties?'

Evaluation

Metrics

Label Accuracy
all 0.8667

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("Are there any sarees with Fekwa Weave technique?")

Training Details

Training Set Metrics

Training set Min Median Max
Word count 4 11.1737 28
Label Training Sample Count
Out of Scope 35
general faq 24
order tracking 34
product discoverability 40
product faq 40
product policy 40

Training Hyperparameters

  • batch_size: (16, 16)
  • num_epochs: (2, 2)
  • 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
0.0004 1 0.256 -
0.0213 50 0.2639 -
0.0425 100 0.2341 -
0.0638 150 0.0407 -
0.0851 200 0.0698 -
0.1063 250 0.014 -
0.1276 300 0.0069 -
0.1489 350 0.0099 -
0.1701 400 0.0014 -
0.1914 450 0.0007 -
0.2127 500 0.0006 -
0.2339 550 0.0005 -
0.2552 600 0.0006 -
0.2765 650 0.0005 -
0.2977 700 0.0002 -
0.3190 750 0.0005 -
0.3403 800 0.0003 -
0.3615 850 0.0003 -
0.3828 900 0.0002 -
0.4041 950 0.0003 -
0.4254 1000 0.0002 -
0.4466 1050 0.0002 -
0.4679 1100 0.0001 -
0.4892 1150 0.0002 -
0.5104 1200 0.0002 -
0.5317 1250 0.0001 -
0.5530 1300 0.0002 -
0.5742 1350 0.0002 -
0.5955 1400 0.0001 -
0.6168 1450 0.0002 -
0.6380 1500 0.0002 -
0.6593 1550 0.0001 -
0.6806 1600 0.0001 -
0.7018 1650 0.0001 -
0.7231 1700 0.0001 -
0.7444 1750 0.0001 -
0.7656 1800 0.0001 -
0.7869 1850 0.0001 -
0.8082 1900 0.0001 -
0.8294 1950 0.0001 -
0.8507 2000 0.0001 -
0.8720 2050 0.0001 -
0.8932 2100 0.0001 -
0.9145 2150 0.0002 -
0.9358 2200 0.0002 -
0.9570 2250 0.0002 -
0.9783 2300 0.0001 -
0.9996 2350 0.0001 -
1.0208 2400 0.0001 -
1.0421 2450 0.0002 -
1.0634 2500 0.0001 -
1.0846 2550 0.0001 -
1.1059 2600 0.0001 -
1.1272 2650 0.0002 -
1.1484 2700 0.0001 -
1.1697 2750 0.0001 -
1.1910 2800 0.0001 -
1.2123 2850 0.0001 -
1.2335 2900 0.0001 -
1.2548 2950 0.0001 -
1.2761 3000 0.0001 -
1.2973 3050 0.0001 -
1.3186 3100 0.0001 -
1.3399 3150 0.0001 -
1.3611 3200 0.0001 -
1.3824 3250 0.0001 -
1.4037 3300 0.0001 -
1.4249 3350 0.0001 -
1.4462 3400 0.0001 -
1.4675 3450 0.0001 -
1.4887 3500 0.0001 -
1.5100 3550 0.0001 -
1.5313 3600 0.0001 -
1.5525 3650 0.0001 -
1.5738 3700 0.0001 -
1.5951 3750 0.0001 -
1.6163 3800 0.0001 -
1.6376 3850 0.0 -
1.6589 3900 0.0001 -
1.6801 3950 0.0001 -
1.7014 4000 0.0001 -
1.7227 4050 0.0001 -
1.7439 4100 0.0001 -
1.7652 4150 0.0001 -
1.7865 4200 0.0001 -
1.8077 4250 0.0001 -
1.8290 4300 0.0001 -
1.8503 4350 0.0001 -
1.8715 4400 0.0 -
1.8928 4450 0.0001 -
1.9141 4500 0.0001 -
1.9353 4550 0.0001 -
1.9566 4600 0.0001 -
1.9779 4650 0.0001 -
1.9991 4700 0.0001 -

Framework Versions

  • Python: 3.10.16
  • SetFit: 1.0.3
  • Sentence Transformers: 2.7.0
  • Transformers: 4.40.2
  • PyTorch: 2.2.2
  • Datasets: 2.19.1
  • 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}
}
Downloads last month
8
Safetensors
Model size
109M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for Shankhdhar/classifier_woog_base_oos

Finetuned
(270)
this model

Evaluation results