fakeBert
This model is a fine-tuned version of bert-base-uncased on a news dataset from Kaggle.
Model description
Fine-tuning Bert for text classification.
Training and evaluation data
Training & Validation: Fake and real news dataset Testing: Fake News Detection Challenge KDD 2020
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-5
- train_batch_size: 16
- eval_batch_size: 16
- optimizer: AdamW