toxicity-score-multi-classification

This model is a fine-tuned version of line-corporation/line-distilbert-base-japanese on a Japanese toxicity dataset. It achieves the following results on the evaluation set:

• Loss: 0.2649
• Roc Auc: 0.7992

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: 8.133692392125703e-06
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Roc Auc
No log	1.0	20	0.6213	0.5148
No log	2.0	40	0.4762	0.4616
No log	3.0	60	0.3754	0.5830
No log	4.0	80	0.3314	0.5706
No log	5.0	100	0.3140	0.5740
No log	6.0	120	0.3067	0.6238
No log	7.0	140	0.3010	0.6645
No log	8.0	160	0.2975	0.7177
No log	9.0	180	0.2949	0.7392
No log	10.0	200	0.2892	0.7731
No log	11.0	220	0.2828	0.7954
No log	12.0	240	0.2827	0.7932
No log	13.0	260	0.2756	0.7984
No log	14.0	280	0.2715	0.8052
No log	15.0	300	0.2733	0.8100
No log	16.0	320	0.2754	0.8142
No log	17.0	340	0.2668	0.8130
No log	18.0	360	0.2642	0.8138
No log	19.0	380	0.2639	0.8117
No log	20.0	400	0.2659	0.8052
No log	21.0	420	0.2646	0.8082
No log	22.0	440	0.2643	0.8039
No log	23.0	460	0.2646	0.8022
No log	24.0	480	0.2644	0.8044
0.2305	25.0	500	0.2639	0.8035
0.2305	26.0	520	0.2639	0.8027
0.2305	27.0	540	0.2647	0.8001
0.2305	28.0	560	0.2643	0.8005
0.2305	29.0	580	0.2649	0.8001
0.2305	30.0	600	0.2649	0.7992

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1
• Datasets 2.14.5
• Tokenizers 0.14.0