metadata

library_name: setfit
tags:
  - setfit
  - sentence-transformers
  - text-classification
  - generated_from_setfit_trainer
metrics:
  - accuracy
widget:
  - text: Ruukki Group calculates that it has lost EUR 4mn in the failed project .
  - text: >-
      The Tecnomen Convergent Charging solution includes functionality for
      prepaid and post-paid billing , charging and rating of voice calls , video
      calls , raw data traffic and any type of content services in both mobile
      and fixed networks .
  - text: The combined value of the planned investments is about EUR 30mn .
  - text: >-
      The Diameter Protocol is developed according to the standards IETF RFC
      3588 and IETF RFC 3539 .
  - text: >-
      Below are unaudited consolidated results for Aspocomp Group under IFRS
      reporting standards .
pipeline_tag: text-classification
inference: true
base_model: BAAI/bge-small-en-v1.5
model-index:
  - name: SetFit with BAAI/bge-small-en-v1.5
    results:
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: Unknown
          type: unknown
          split: test
        metrics:
          - type: accuracy
            value: 0.9426048565121413
            name: Accuracy

SetFit with BAAI/bge-small-en-v1.5

This is a SetFit model that can be used for Text Classification. This SetFit model uses BAAI/bge-small-en-v1.5 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: BAAI/bge-small-en-v1.5
Classification head: a LogisticRegression instance
Maximum Sequence Length: 512 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
positive	'HELSINKI ( AFX ) - Nokian Tyres reported a fourth quarter pretax profit of 61.5 mln eur , up from 48.6 mln on the back of strong sales .' 'Equity ratio was 60.9 % compared to 54.2 % In the third quarter of 2007 , net sales of the Frozen Foods Business totaled EUR 11.0 , up by about 5 % from the third quarter of 2006 .' "`` After a long , unprofitable period the Food Division posted a profitable result , which speaks of a healthier cost structure and a new approach in business operations , '' Rihko said ."
neutral	'Their names have not yet been released .' 'The contract includes design , construction , delivery of equipment , installation and commissioning .' "Tieto 's service is also used to send , process and receive materials related to absentee voting ."
negative	'The company confirmed its estimate for lower revenue for the whole 2009 than the year-ago EUR93 .9 m as given in the interim report on 5 August 2009 .' 'Acando AB ( ACANB SS ) fell 8.9 percent to 13.35 kronor , the lowest close since Dec. 11 .' 'Okmetic expects its net sales for the first half of 2009 to be less than in 2008 .'

Evaluation

Metrics

Label	Accuracy
all	0.9426

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("moshew/bge-small-en-v1.5-SetFit-FSA")
# Run inference
preds = model("The combined value of the planned investments is about EUR 30mn .")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	2	22.4020	60

Label	Training Sample Count
negative	266
neutral	1142
positive	403

Training Hyperparameters

batch_size: (16, 16)
num_epochs: (1, 1)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 10
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
0.0004	1	0.2832	-
0.0221	50	0.209	-
0.0442	100	0.1899	-
0.0663	150	0.1399	-
0.0883	200	0.1274	-
0.1104	250	0.0586	-
0.1325	300	0.0756	-
0.1546	350	0.0777	-
0.1767	400	0.0684	-
0.1988	450	0.0311	-
0.2208	500	0.0102	-
0.2429	550	0.052	-
0.2650	600	0.0149	-
0.2871	650	0.1042	-
0.3092	700	0.061	-
0.3313	750	0.0083	-
0.3534	800	0.0036	-
0.3754	850	0.002	-
0.3975	900	0.0598	-
0.4196	950	0.0036	-
0.4417	1000	0.0027	-
0.4638	1050	0.0617	-
0.4859	1100	0.0015	-
0.5080	1150	0.0022	-
0.5300	1200	0.0016	-
0.5521	1250	0.0009	-
0.5742	1300	0.0013	-
0.5963	1350	0.0009	-
0.6184	1400	0.0015	-
0.6405	1450	0.0018	-
0.6625	1500	0.0015	-
0.6846	1550	0.0018	-
0.7067	1600	0.0016	-
0.7288	1650	0.0022	-
0.7509	1700	0.0013	-
0.7730	1750	0.0108	-
0.7951	1800	0.0016	-
0.8171	1850	0.0021	-
0.8392	1900	0.002	-
0.8613	1950	0.0015	-
0.8834	2000	0.0016	-
0.9055	2050	0.0028	-
0.9276	2100	0.0013	-
0.9496	2150	0.0019	-
0.9717	2200	0.0075	-
0.9938	2250	0.0015	-

Framework Versions

Python: 3.10.12
SetFit: 1.0.3
Sentence Transformers: 2.5.1
Transformers: 4.38.1
PyTorch: 2.1.0+cu121
Datasets: 2.18.0
Tokenizers: 0.15.2

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}
}