HylkeBr/bertje_dutch-cola

This model is a fine-tuned version of GroNLP/bert-base-dutch-cased on GroNLP/dutch-cola, from where only the acceptable and unacceptable examples were used to fine-tune. The data used for fine-tuning therefore only contains sentences with original mark 'None' and '*'.

Accuracy: 0.8127

The following hyperparameters were used during training:

learning rate: 4e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
num_epochs: 4

Epoch	Training Loss	Step	Validation Loss	Accuracy
1.0	0.4971	1169	0.4629	0.7914
2.0	0.2478	2338	0.7041	0.7814
3.0	0.1333	3507	0.9777	0.8064
4.0	0.0687	4676	1.1322	0.8127