donut-base-hangul-handwritten-KMOU

This model is a fine-tuned version of naver-clova-ix/donut-base on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.0000

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: 2e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
3.5986	1.0	157	2.9011
0.6912	2.0	314	0.4334
0.3317	3.0	471	0.2250
0.2029	4.0	628	0.1379
0.1587	5.0	785	0.0901
0.0963	6.0	942	0.0609
0.0812	7.0	1099	0.0410
0.0755	8.0	1256	0.0289
0.0471	9.0	1413	0.0191
0.0368	10.0	1570	0.0140
0.0247	11.0	1727	0.0088
0.023	12.0	1884	0.0065
0.0239	13.0	2041	0.0045
0.0139	14.0	2198	0.0036
0.0166	15.0	2355	0.0025
0.0144	16.0	2512	0.0020
0.0144	17.0	2669	0.0017
0.012	18.0	2826	0.0012
0.0085	19.0	2983	0.0008
0.0063	20.0	3140	0.0010
0.0054	21.0	3297	0.0007
0.0074	22.0	3454	0.0008
0.0056	23.0	3611	0.0006
0.0048	24.0	3768	0.0004
0.0055	25.0	3925	0.0003
0.0053	26.0	4082	0.0003
0.0054	27.0	4239	0.0003
0.0067	28.0	4396	0.0004
0.0042	29.0	4553	0.0002
0.0028	30.0	4710	0.0003
0.0036	31.0	4867	0.0002
0.004	32.0	5024	0.0003
0.0041	33.0	5181	0.0001
0.005	34.0	5338	0.0002
0.0025	35.0	5495	0.0002
0.004	36.0	5652	0.0001
0.0038	37.0	5809	0.0001
0.0033	38.0	5966	0.0001
0.0026	39.0	6123	0.0001
0.0052	40.0	6280	0.0001
0.0051	41.0	6437	0.0001
0.0019	42.0	6594	0.0001
0.0036	43.0	6751	0.0001
0.0053	44.0	6908	0.0001
0.0018	45.0	7065	0.0001
0.0033	46.0	7222	0.0001
0.0061	47.0	7379	0.0001
0.0023	48.0	7536	0.0000
0.0023	49.0	7693	0.0000
0.0048	50.0	7850	0.0000

Framework versions

• Transformers 4.28.1
• Pytorch 1.10.0+cu111
• Datasets 2.11.0
• Tokenizers 0.13.3