VaccinChatSentenceClassifierDutch_fromBERTje

This model is a fine-tuned version of GroNLP/bert-base-dutch-cased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6223
• Accuracy: 0.9068

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.4666	1.0	1320	2.3355	0.5768
1.5293	2.0	2640	1.1118	0.8144
0.8031	3.0	3960	0.6362	0.8803
0.2985	4.0	5280	0.5119	0.8958
0.1284	5.0	6600	0.5023	0.8931
0.0842	6.0	7920	0.5246	0.9022
0.0414	7.0	9240	0.5581	0.9013
0.0372	8.0	10560	0.5721	0.9004
0.0292	9.0	11880	0.5469	0.9141
0.0257	10.0	13200	0.5871	0.9059
0.0189	11.0	14520	0.6181	0.9049
0.0104	12.0	15840	0.6184	0.9068
0.009	13.0	17160	0.6013	0.9049
0.0051	14.0	18480	0.6205	0.9059
0.0035	15.0	19800	0.6223	0.9068

Framework versions

• Transformers 4.13.0.dev0
• Pytorch 1.10.0
• Datasets 1.16.1
• Tokenizers 0.10.3