0fce00dd8ceb0cee0320d0cd81ebfdc4

This model is a fine-tuned version of studio-ousia/luke-japanese-base on the dair-ai/emotion [split] dataset. It achieves the following results on the evaluation set:

• Loss: 0.3045
• Data Size: 1.0
• Epoch Runtime: 45.8828
• Accuracy: 0.9103
• F1 Macro: 0.8702

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
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	1.7667	0	2.3001	0.2908	0.0751
No log	1	500	1.5884	0.0078	2.8614	0.3488	0.0862
No log	2	1000	1.5733	0.0156	3.2727	0.3488	0.0862
No log	3	1500	1.5926	0.0312	4.1961	0.3488	0.0862
No log	4	2000	1.5815	0.0625	5.8628	0.3236	0.1260
0.0859	5	2500	1.5851	0.125	8.4869	0.2908	0.0751
1.5604	6	3000	1.4602	0.25	14.1540	0.4541	0.1822
0.1819	7	3500	0.8192	0.5	24.8717	0.7278	0.5679
0.4683	8.0	4000	0.3858	1.0	46.2410	0.8705	0.8248
0.2829	9.0	4500	0.3304	1.0	45.9392	0.8871	0.8374
0.268	10.0	5000	0.2746	1.0	45.2703	0.9012	0.8548
0.2017	11.0	5500	0.3066	1.0	45.4689	0.8962	0.8549
0.1925	12.0	6000	0.2728	1.0	45.3819	0.9073	0.8619
0.1847	13.0	6500	0.2389	1.0	45.0696	0.9012	0.8674
0.1541	14.0	7000	0.2594	1.0	46.3390	0.9103	0.8647
0.1706	15.0	7500	0.2634	1.0	45.9249	0.9103	0.8636
0.1346	16.0	8000	0.2650	1.0	46.2234	0.9078	0.8713
0.1241	17.0	8500	0.3045	1.0	45.8828	0.9103	0.8702

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1