trocr-small-printedkorean-deleteunusedchar_noise

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.3375
• Cer: 0.2783
• Wer: 0.2975
• Accuracy: 45.6667

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: 4e-05
• train_batch_size: 128
• eval_batch_size: 192
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer	Accuracy
1.711	0.43	1000	1.6485	0.3288	0.3944	30.6667
1.6849	0.85	2000	1.5361	0.3098	0.3809	32.3333
1.4933	1.28	3000	1.4302	0.2935	0.3533	34.6667
1.526	1.71	4000	1.4010	0.2922	0.3400	35.6667
1.3422	2.13	5000	1.3883	0.2846	0.3331	36.0
1.333	2.56	6000	1.3790	0.2871	0.3308	34.0
1.3295	2.99	7000	1.3644	0.2876	0.3294	35.6667
1.3294	3.42	8000	1.3588	0.2824	0.3202	36.6667
1.3578	3.84	9000	1.3502	0.2823	0.3162	40.6667
1.3029	4.27	10000	1.3514	0.2879	0.3228	37.0
1.2777	4.7	11000	1.3507	0.2813	0.3168	38.3333
1.1781	5.12	12000	1.3507	0.2791	0.3150	40.3333
1.3025	5.55	13000	1.3459	0.2818	0.3099	41.6667
1.2024	5.98	14000	1.3401	0.2801	0.3061	41.6667
1.1792	6.4	15000	1.3412	0.2763	0.3015	44.6667
1.1586	6.83	16000	1.3410	0.2799	0.3064	43.3333
1.2098	7.26	17000	1.3439	0.2777	0.3030	43.6667
1.2122	7.69	18000	1.3418	0.2816	0.3050	43.3333
1.1323	8.11	19000	1.3409	0.2767	0.2981	45.3333
1.2215	8.54	20000	1.3386	0.2781	0.3004	44.0
1.2068	8.97	21000	1.3375	0.2762	0.2972	45.0
1.0847	9.39	22000	1.3366	0.2765	0.2969	46.0
1.1791	9.82	23000	1.3375	0.2783	0.2975	45.6667

Framework versions

• Transformers 4.28.0
• Pytorch 1.13.1+cu116
• Datasets 2.14.4
• Tokenizers 0.13.3