classification-hate-speech-DE-14-twit

This model is a fine-tuned version of indolem/indobertweet-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.0647
• F1 macro: 0.4076
• Weighted: 0.5816
• Balanced accuracy: 0.5560

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: 5e-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: 14

Training results

Training Loss	Epoch	Step	Validation Loss	F1 macro	Weighted	Balanced accuracy
1.2717	1.0	152	1.1413	0.3986	0.6607	0.4780
0.7363	2.0	304	1.2159	0.3829	0.6075	0.4706
0.2523	3.0	456	1.4802	0.3994	0.6380	0.4901
0.1209	4.0	608	1.9894	0.4200	0.6130	0.5900
0.0201	5.0	760	2.7075	0.3731	0.5537	0.5221
0.0052	6.0	912	2.4151	0.4301	0.6211	0.5905
0.0012	7.0	1064	3.2430	0.3791	0.5228	0.5470
0.0014	8.0	1216	2.7395	0.4054	0.6010	0.5485
0.001	9.0	1368	2.6392	0.4167	0.6052	0.5505
0.0005	10.0	1520	2.8755	0.4049	0.5893	0.5546
0.0005	11.0	1672	2.7610	0.4152	0.6027	0.5560
0.0009	12.0	1824	3.0945	0.4011	0.5710	0.5511
0.0003	13.0	1976	3.1262	0.4003	0.5687	0.5506
0.0004	14.0	2128	3.0647	0.4076	0.5816	0.5560

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1