SetFit with BAAI/bge-large-en-v1.5
This is a SetFit model that can be used for Text Classification. This SetFit model uses BAAI/bge-large-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 Sources
Model Labels
Label |
Examples |
peak |
- 'after using product to summarize and gather main points of hundreds of research articles that are 50+ pages, i think i can confidently say that brand is on the right track with regards to implementing product in their business. truly extraordinary.'
- 'i was stuck in a error for 2+ hours and my bingey bot cleared it!! awesome ai product'
- 'product in teams: in teams, product transforms meetings. it organizes thoughts, maintains context, and facilitates collaborative brainstorming, making every meeting more productive.'
|
neither |
- ">youll receive the test via email and will have two hours to complete it. finally, youll return to zoom with the analyst to go over your results together i don't think it's live. op will get the assigment and he/she has 2 hours to complete it. if this is correct, then op is an idiot because there are thousands of examples online and then there's product. op, start working on the fundamentals and pay the $20 product suscription for product."
- 'utilising advanced technologies with brand to perform a practical demonstration for a client on themes of cyber security, product, product, digital transformation, product, the product and more. these skills are rapidly being adopted for safety and efficielnkd.in/ghumbffm'
- "another great example of the elites in the tech world using control of the information to infl your thoughts and actions. as product becomes more prevalent doing your own research will be essential. will be interesting to see if anyone finds success with designing a true 'unbiased' product"
|
pit |
- "the utter disappointment of learning from an amazing passionate teacher for two years who gives you decades of knowledge in 2 years and then you continue the subject and get some bland intellectual from the capital who can't even make a product presentation"
- 'the amount of times that product has been forced on me against my will after updates is just infuriating. product just taking advantage of the market position they (illegally) established long ago. near-universal software compatibility and being the default os of the general market are why people keep using them. they are in the position where they can fail upwards. and it sucks for the rest of us.'
- 'literally canceling my subscription on my product because this is terrible business practice. forcing subscription services to squeeze out every last dollar is disgusting especially when your whole program is a rip off of another established program. cringe'
|
Evaluation
Metrics
Label |
Accuracy |
F1 |
Precision |
Recall |
all |
0.88 |
[0.8846153846153847, 0.6666666666666666, 0.9222520107238605] |
[0.8214285714285714, 0.5, 1.0] |
[0.9583333333333334, 1.0, 0.8557213930348259] |
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
model = SetFitModel.from_pretrained("jamiehudson/725_model_v6")
preds = model("why though? whats the harm in using ai as a tool. theres more to ai than product.")
Training Details
Training Set Metrics
Training set |
Min |
Median |
Max |
Word count |
10 |
37.08 |
98 |
Label |
Training Sample Count |
pit |
50 |
peak |
50 |
neither |
50 |
Training Hyperparameters
- batch_size: (16, 16)
- num_epochs: (3, 3)
- 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: False
Training Results
Epoch |
Step |
Training Loss |
Validation Loss |
0.0011 |
1 |
0.2299 |
- |
0.0533 |
50 |
0.1604 |
- |
0.1066 |
100 |
0.0071 |
- |
0.1599 |
150 |
0.0016 |
- |
0.2132 |
200 |
0.0012 |
- |
0.2665 |
250 |
0.0012 |
- |
0.3198 |
300 |
0.0011 |
- |
0.3731 |
350 |
0.0009 |
- |
0.4264 |
400 |
0.0008 |
- |
0.4797 |
450 |
0.0009 |
- |
0.5330 |
500 |
0.0007 |
- |
0.5864 |
550 |
0.0008 |
- |
0.6397 |
600 |
0.0007 |
- |
0.6930 |
650 |
0.0007 |
- |
0.7463 |
700 |
0.0007 |
- |
0.7996 |
750 |
0.0006 |
- |
0.8529 |
800 |
0.0006 |
- |
0.9062 |
850 |
0.0006 |
- |
0.9595 |
900 |
0.0006 |
- |
0.0011 |
1 |
0.0006 |
- |
0.0533 |
50 |
0.0005 |
- |
0.1066 |
100 |
0.0005 |
- |
0.1599 |
150 |
0.0005 |
- |
0.2132 |
200 |
0.0004 |
- |
0.2665 |
250 |
0.0003 |
- |
0.3198 |
300 |
0.0004 |
- |
0.3731 |
350 |
0.0003 |
- |
0.4264 |
400 |
0.0004 |
- |
0.4797 |
450 |
0.0004 |
- |
0.5330 |
500 |
0.0002 |
- |
0.5864 |
550 |
0.0002 |
- |
0.6397 |
600 |
0.0002 |
- |
0.6930 |
650 |
0.0002 |
- |
0.7463 |
700 |
0.0002 |
- |
0.7996 |
750 |
0.0003 |
- |
0.8529 |
800 |
0.0002 |
- |
0.9062 |
850 |
0.0002 |
- |
0.9595 |
900 |
0.0001 |
- |
1.0128 |
950 |
0.0002 |
- |
1.0661 |
1000 |
0.0002 |
- |
1.1194 |
1050 |
0.0002 |
- |
1.1727 |
1100 |
0.0001 |
- |
1.2260 |
1150 |
0.0001 |
- |
1.2793 |
1200 |
0.0001 |
- |
1.3326 |
1250 |
0.0001 |
- |
1.3859 |
1300 |
0.0001 |
- |
1.4392 |
1350 |
0.0001 |
- |
1.4925 |
1400 |
0.0001 |
- |
1.5458 |
1450 |
0.0001 |
- |
1.5991 |
1500 |
0.0001 |
- |
1.6525 |
1550 |
0.0001 |
- |
1.7058 |
1600 |
0.0001 |
- |
1.7591 |
1650 |
0.0001 |
- |
1.8124 |
1700 |
0.0001 |
- |
1.8657 |
1750 |
0.0001 |
- |
1.9190 |
1800 |
0.0001 |
- |
1.9723 |
1850 |
0.0001 |
- |
2.0256 |
1900 |
0.0001 |
- |
2.0789 |
1950 |
0.0001 |
- |
2.1322 |
2000 |
0.0001 |
- |
2.1855 |
2050 |
0.0001 |
- |
2.2388 |
2100 |
0.0001 |
- |
2.2921 |
2150 |
0.0001 |
- |
2.3454 |
2200 |
0.0001 |
- |
2.3987 |
2250 |
0.0001 |
- |
2.4520 |
2300 |
0.0001 |
- |
2.5053 |
2350 |
0.0001 |
- |
2.5586 |
2400 |
0.0001 |
- |
2.6119 |
2450 |
0.0001 |
- |
2.6652 |
2500 |
0.0001 |
- |
2.7186 |
2550 |
0.0001 |
- |
2.7719 |
2600 |
0.0001 |
- |
2.8252 |
2650 |
0.0001 |
- |
2.8785 |
2700 |
0.0001 |
- |
2.9318 |
2750 |
0.0001 |
- |
2.9851 |
2800 |
0.0001 |
- |
Framework Versions
- Python: 3.10.12
- SetFit: 1.0.3
- Sentence Transformers: 2.5.1
- Transformers: 4.38.2
- 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}
}