unsmile_classifier_v1

This model is a fine-tuned version of monologg/koelectra-base-v3-discriminator on the smilegate-ai/kor_unsmile dataset. It achieves the following results on the evaluation set:

• Loss: 0.1510
• Toxic Macro F1: 0.7069
• Toxic Macro Precision: 0.7169
• Toxic Macro Recall: 0.6983

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: 2.907776989257978e-05
• train_batch_size: 8
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Toxic Macro F1	Toxic Macro Precision	Toxic Macro Recall
0.2589	1.0	1876	0.1953	0.5538	0.6405	0.5330
0.1518	2.0	3752	0.1503	0.6958	0.6986	0.6951
0.1101	3.0	5628	0.1510	0.7069	0.7169	0.6983

Framework versions

• Transformers 4.49.0
• Pytorch 2.11.0+cu128
• Datasets 3.3.2
• Tokenizers 0.21.4