metadata

library_name: setfit
tags:
  - setfit
  - sentence-transformers
  - text-classification
  - generated_from_setfit_trainer
datasets:
  - dvilasuero/banking77-topics-setfit
metrics:
  - accuracy
widget:
  - text: I requested a refund, and never received it. What can I do?
  - text: I have a 1 euro fee on my statement.
  - text: I would like an account for my children, how do I go about doing this?
  - text: What do I need to do to transfer money into my account?
  - text: Which country's currency do you support?
pipeline_tag: text-classification
inference: true
base_model: thenlper/gte-large
model-index:
  - name: SetFit with thenlper/gte-large
    results:
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: dvilasuero/banking77-topics-setfit
          type: dvilasuero/banking77-topics-setfit
          split: test
        metrics:
          - type: accuracy
            value: 0.9230769230769231
            name: Accuracy

SetFit with thenlper/gte-large

This is a SetFit model trained on the dvilasuero/banking77-topics-setfit dataset that can be used for Text Classification. This SetFit model uses thenlper/gte-large 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: thenlper/gte-large
Classification head: a LogisticRegression instance
Maximum Sequence Length: 512 tokens
Number of Classes: 8 classes
Training Dataset: dvilasuero/banking77-topics-setfit

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
2	'The money I transferred does not show in the balance.' 'I was wondering how I could have two charges for the same item happen more than once in a 7 day period. Is there anyway I could get this corrected asap.' 'What is the source of my available funds?'
0	'Do you support the EU?' "Can you freeze my account? I just saw there are transactions on my account that I don't recognize. How can I fix this?" 'Please close my account. I am unsatisfied with your service.'
5	'Are you able to unblock my pin?' 'I can not find my card pin.' 'If I need a PIN for my card, where is it located?'
1	"I can't get money out of the ATM" 'Where can I use this card at an ATM?' 'Can I use my card at any ATMs?'
3	'Can I get cash with this card anywhere?' 'Can you please show me where I can find the location to link my card?' 'Am I able to get a card in EU?'
6	'My friends want to top up my account' 'Can I be topped up once I hit a certain balance?' 'Can you tell me why my top up was reverted?'
7	'How do I send my account money through transfer?' 'How do I transfer money to my account?' 'How can I transfer money from an outside bank?'
4	'Do you work with all fiat currencies?' 'Can I exchange to EUR?' 'Is my country supported'

Evaluation

Metrics

Label	Accuracy
all	0.9231

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("HarshalBhg/gte-large-setfit-train-test2")
# Run inference
preds = model("I have a 1 euro fee on my statement.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	4	10.5833	40

Label	Training Sample Count
0	10
1	19
2	28
3	36
4	13
5	14
6	15
7	21

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (1, 1)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 20
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: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0026	1	0.3183	-
0.1282	50	0.0614	-
0.2564	100	0.0044	-
0.3846	150	0.001	-
0.5128	200	0.0008	-
0.6410	250	0.001	-
0.7692	300	0.0006	-
0.8974	350	0.0012	-

Framework Versions

Python: 3.10.12
SetFit: 1.0.1
Sentence Transformers: 2.2.2
Transformers: 4.35.2
PyTorch: 2.1.0+cu121
Datasets: 2.15.0
Tokenizers: 0.15.0

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}
}