metadata
license: apache-2.0
tags:
- generated_from_keras_callback
model-index:
- name: athrado/bert-finetuned-nli
results: []
athrado/bert-finetuned-nli
This model is a fine-tuned version of bert-base-uncased on the sick dataset for natural language inference. It achieves the following results on the evaluation set:
- Train Loss: 0.0671
- Train Accuracy: 0.9806
- Validation Loss: 0.5285
- Validation Accuracy: 0.8424
- Epoch: 4
Model description
Example model for educational purposes: fine-tuning the bert-base-uncased model for natural language inference.
Intended uses & limitations
- Learning about transformer model and fine-tuning
- Natural language inference fine-tuned on small dataset
Training and evaluation data
The model is evaluated using the sick validation. We report accuracy, and in addition we computed a weighted F1-score of 0.842.
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': False, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 5e-05, 'decay_steps': 2775, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
- training_precision: float32
Training results
| Train Loss | Train Accuracy | Validation Loss | Validation Accuracy | Epoch |
|---|---|---|---|---|
| 0.5345 | 0.7810 | 0.4586 | 0.8283 | 0 |
| 0.3253 | 0.8797 | 0.3890 | 0.8404 | 1 |
| 0.2070 | 0.9290 | 0.4210 | 0.8303 | 2 |
| 0.1171 | 0.9610 | 0.5200 | 0.8424 | 3 |
| 0.0671 | 0.9806 | 0.5285 | 0.8424 | 4 |
Framework versions
- Transformers 4.30.2
- TensorFlow 2.12.0
- Datasets 2.13.1
- Tokenizers 0.13.3