koelectra_emotion_v3

This model is a fine-tuned version of monologg/koelectra-small-finetuned-sentiment on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.5162
• Accuracy: 0.4089
• F1: 0.4177

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: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
1.8421	1.0	3723	1.7489	0.2754	0.2071
1.6975	2.0	7446	1.6315	0.3388	0.3192
1.6025	3.0	11169	1.5650	0.3819	0.3749
1.5331	4.0	14892	1.5103	0.4024	0.3932
1.4824	5.0	18615	1.4976	0.4120	0.4063
1.4378	6.0	22338	1.4715	0.4249	0.4243
1.393	7.0	26061	1.4635	0.4313	0.4309
1.3493	8.0	29784	1.4839	0.4304	0.4288
1.3069	9.0	33507	1.4705	0.4342	0.4360
1.2639	10.0	37230	1.4816	0.4363	0.4409
1.2218	11.0	40953	1.5249	0.4311	0.4352
1.1793	12.0	44676	1.5316	0.4309	0.4372
1.1342	13.0	48399	1.5984	0.4302	0.4355
1.0891	14.0	52122	1.6145	0.4265	0.4329
1.0445	15.0	55845	1.6517	0.4273	0.4354
0.9999	16.0	59568	1.6906	0.4266	0.4307
0.9593	17.0	63291	1.7398	0.4217	0.4297
0.9134	18.0	67014	1.8011	0.4162	0.4270
0.8697	19.0	70737	1.8496	0.4200	0.4262
0.8281	20.0	74460	1.8960	0.4169	0.4241
0.7899	21.0	78183	1.9289	0.4157	0.4246
0.7519	22.0	81906	2.0116	0.4166	0.4243
0.7125	23.0	85629	2.0726	0.4148	0.4226
0.6757	24.0	89352	2.1433	0.4081	0.4179
0.6417	25.0	93075	2.1700	0.4025	0.4160
0.6139	26.0	96798	2.2486	0.4061	0.4159
0.5799	27.0	100521	2.3329	0.4051	0.4155
0.5529	28.0	104244	2.3371	0.4102	0.4174
0.5229	29.0	107967	2.4037	0.4021	0.4151
0.4992	30.0	111690	2.5162	0.4089	0.4177

Framework versions

• Transformers 4.56.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.0