text stringclasses 9
values | label stringclasses 3
values |
|---|---|
Superb craftsmanship and design | positive |
Failed to meet advertised claims | negative |
Reasonable quality for budget option | neutral |
Best in class, truly exceptional | positive |
Arrived late and missing parts | negative |
Fair product with minor issues | neutral |
Top-notch, will definitely buy again | positive |
Horrible, returned immediately | negative |
Acceptable but room for improvement | neutral |
BestRunClassifier Dataset
This dataset contains the training and validation splits from the best-performing hyperparameter run of our text classifier.
Run Configuration
| Parameter | Value |
|---|---|
| Learning Rate | 0.0003 |
| Batch Size | 32 |
| Epochs | 20 |
| Dropout | 0.3 |
Metrics
| Metric | Value |
|---|---|
| Train Loss | 0.195 |
| Val Loss | 0.342 |
| Train Accuracy | 0.941 |
| Val Accuracy | 0.872 |
| Train F1 | 0.938 |
| Val F1 | 0.847 |
Dataset Structure
train.csv: Training split with columnstextandlabel(positive/negative/neutral).val.csv: Validation split with columnstextandlabel(positive/negative/neutral).metrics.json: Full metrics and configuration for the best run.
Usage
from datasets import load_dataset
ds = load_dataset("username/BestRunClassifier-Data")
License
This dataset is released under the MIT License.
- Downloads last month
- 36