metadata
base_model: akhooli/sbert_ar_nli_500k_norm
library_name: setfit
metrics:
- accuracy
pipeline_tag: text-classification
tags:
- setfit
- sentence-transformers
- text-classification
- generated_from_setfit_trainer
widget:
- text: >-
يا زلمة يلي بيصنع معنا معروف بنتشكره شو ما كان يكون وانتم ادعياء الاخوة
العرب هول مش ايرانيين ولا عجم عرب متلنا متلهم
- text: لعمي
- text: هلق رجع لمن قلو الريس تبعو هش قلو مشمو على عيني ؟
- text: >-
مثل الكليشيه وبشكل يومي في حدا بده يعاير التاني بيقوم بيشبهه بالكلب والله
اذا حدا شبهني بالكلب بعتبرها مدح شديد
- text: الله لا يحرمك من الهبل ان شاء الله
inference: true
model-index:
- name: SetFit with akhooli/sbert_ar_nli_500k_norm
results:
- task:
type: text-classification
name: Text Classification
dataset:
name: Unknown
type: unknown
split: test
metrics:
- type: accuracy
value: 0.8497652582159625
name: Accuracy
SetFit with akhooli/sbert_ar_nli_500k_norm
This is a SetFit model that can be used for Text Classification. This SetFit model uses akhooli/sbert_ar_nli_500k_norm 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: akhooli/sbert_ar_nli_500k_norm
- Classification head: a LogisticRegression instance
- Maximum Sequence Length: 512 tokens
- Number of Classes: 2 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 |
---|---|
negative |
|
positive |
|
Evaluation
Metrics
Label | Accuracy |
---|---|
all | 0.8498 |
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("akhooli/setfit_ar_hs")
# Run inference
preds = model("لعمي")
Training Details
Training Set Metrics
Training set | Min | Median | Max |
---|---|---|---|
Word count | 1 | 12.2323 | 52 |
Label | Training Sample Count |
---|---|
negative | 1995 |
positive | 2500 |
Training Hyperparameters
- batch_size: (32, 32)
- num_epochs: (1, 1)
- max_steps: 10000
- sampling_strategy: undersampling
- 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
- l2_weight: 0.01
- seed: 42
- run_name: setfit_hate_25kv
- eval_max_steps: -1
- load_best_model_at_end: False
Training Results
Epoch | Step | Training Loss | Validation Loss |
---|---|---|---|
0.0002 | 1 | 0.3185 | - |
0.02 | 100 | 0.2901 | - |
0.04 | 200 | 0.2441 | - |
0.06 | 300 | 0.2209 | - |
0.08 | 400 | 0.1715 | - |
0.1 | 500 | 0.1304 | - |
0.12 | 600 | 0.0891 | - |
0.14 | 700 | 0.0604 | - |
0.16 | 800 | 0.0436 | - |
0.18 | 900 | 0.0408 | - |
0.2 | 1000 | 0.0265 | - |
0.22 | 1100 | 0.0239 | - |
0.24 | 1200 | 0.0235 | - |
0.26 | 1300 | 0.0232 | - |
0.28 | 1400 | 0.0241 | - |
0.3 | 1500 | 0.019 | - |
0.32 | 1600 | 0.0168 | - |
0.34 | 1700 | 0.0172 | - |
0.36 | 1800 | 0.0136 | - |
0.38 | 1900 | 0.0099 | - |
0.4 | 2000 | 0.0117 | - |
0.42 | 2100 | 0.0091 | - |
0.44 | 2200 | 0.0067 | - |
0.46 | 2300 | 0.0074 | - |
0.48 | 2400 | 0.0055 | - |
0.5 | 2500 | 0.0053 | - |
0.52 | 2600 | 0.0054 | - |
0.54 | 2700 | 0.0058 | - |
0.56 | 2800 | 0.0059 | - |
0.58 | 2900 | 0.0055 | - |
0.6 | 3000 | 0.0043 | - |
0.62 | 3100 | 0.0045 | - |
0.64 | 3200 | 0.0055 | - |
0.66 | 3300 | 0.0042 | - |
0.68 | 3400 | 0.0024 | - |
0.7 | 3500 | 0.0025 | - |
0.72 | 3600 | 0.0047 | - |
0.74 | 3700 | 0.0036 | - |
0.76 | 3800 | 0.0029 | - |
0.78 | 3900 | 0.0043 | - |
0.8 | 4000 | 0.0036 | - |
0.82 | 4100 | 0.0025 | - |
0.84 | 4200 | 0.0033 | - |
0.86 | 4300 | 0.0018 | - |
0.88 | 4400 | 0.0016 | - |
0.9 | 4500 | 0.0018 | - |
0.92 | 4600 | 0.0023 | - |
0.94 | 4700 | 0.0027 | - |
0.96 | 4800 | 0.0023 | - |
0.98 | 4900 | 0.0012 | - |
1.0 | 5000 | 0.0021 | - |
1.02 | 5100 | 0.0026 | - |
1.04 | 5200 | 0.0019 | - |
1.06 | 5300 | 0.002 | - |
1.08 | 5400 | 0.0022 | - |
1.1 | 5500 | 0.0025 | - |
1.12 | 5600 | 0.0033 | - |
1.1400 | 5700 | 0.001 | - |
1.16 | 5800 | 0.0016 | - |
1.18 | 5900 | 0.0015 | - |
1.2 | 6000 | 0.0008 | - |
1.22 | 6100 | 0.0011 | - |
1.24 | 6200 | 0.0012 | - |
1.26 | 6300 | 0.0009 | - |
1.28 | 6400 | 0.0012 | - |
1.3 | 6500 | 0.001 | - |
1.32 | 6600 | 0.0014 | - |
1.34 | 6700 | 0.0002 | - |
1.3600 | 6800 | 0.0005 | - |
1.38 | 6900 | 0.0003 | - |
1.4 | 7000 | 0.0001 | - |
1.42 | 7100 | 0.0007 | - |
1.44 | 7200 | 0.0003 | - |
1.46 | 7300 | 0.0002 | - |
1.48 | 7400 | 0.0005 | - |
1.5 | 7500 | 0.0001 | - |
1.52 | 7600 | 0.0003 | - |
1.54 | 7700 | 0.001 | - |
1.56 | 7800 | 0.0003 | - |
1.58 | 7900 | 0.0 | - |
1.6 | 8000 | 0.0002 | - |
1.62 | 8100 | 0.0 | - |
1.6400 | 8200 | 0.0002 | - |
1.6600 | 8300 | 0.0002 | - |
1.6800 | 8400 | 0.0 | - |
1.7 | 8500 | 0.0 | - |
1.72 | 8600 | 0.0002 | - |
1.74 | 8700 | 0.0002 | - |
1.76 | 8800 | 0.0002 | - |
1.78 | 8900 | 0.0002 | - |
1.8 | 9000 | 0.0 | - |
1.8200 | 9100 | 0.0004 | - |
1.8400 | 9200 | 0.0 | - |
1.8600 | 9300 | 0.0002 | - |
1.88 | 9400 | 0.0002 | - |
1.9 | 9500 | 0.0 | - |
1.92 | 9600 | 0.0003 | - |
1.94 | 9700 | 0.0 | - |
1.96 | 9800 | 0.0 | - |
1.98 | 9900 | 0.0 | - |
2.0 | 10000 | 0.0 | - |
Framework Versions
- Python: 3.10.14
- SetFit: 1.2.0.dev0
- Sentence Transformers: 3.2.1
- Transformers: 4.45.1
- PyTorch: 2.4.0
- Datasets: 3.0.1
- Tokenizers: 0.20.0
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}
}