IS557_TrOCR_AllData

This model is a fine-tuned version of microsoft/trocr-base-stage1 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.0124
• Cer: 0.1764

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
4.4421	0.0708	200	4.1713	0.6172
3.2966	0.1415	400	3.1228	0.5064
2.5955	0.2123	600	2.7626	0.3949
2.201	0.2831	800	2.4729	0.4258
1.5876	0.3539	1000	2.2295	0.3097
2.7774	0.4246	1200	2.1112	0.3092
2.7995	0.4954	1400	2.1286	0.3070
1.5241	0.5662	1600	1.8915	0.2534
0.9877	0.6369	1800	1.8091	0.2624
1.21	0.7077	2000	1.7322	0.2552
1.5582	0.7785	2200	1.6645	0.2445
1.1392	0.8493	2400	1.6458	0.2294
1.8181	0.9200	2600	1.5874	0.2313
1.2517	0.9908	2800	1.5736	0.2361
1.2004	1.0616	3000	1.5693	0.2344
1.4357	1.1323	3200	1.5447	0.2190
1.1354	1.2031	3400	1.4577	0.3232
0.853	1.2739	3600	1.4034	0.2166
1.4631	1.3447	3800	1.4105	0.2093
2.3234	1.4154	4000	1.3813	0.2044
0.6976	1.4862	4200	1.3505	0.2088
1.4337	1.5570	4400	1.3338	0.2079
1.0502	1.6277	4600	1.3044	0.1900
1.2216	1.6985	4800	1.2883	0.2176
1.0111	1.7693	5000	1.2530	0.1977
0.9992	1.8401	5200	1.2245	0.2064
0.9941	1.9108	5400	1.2187	0.1933
1.4861	1.9816	5600	1.1961	0.1827
1.1703	2.0524	5800	1.1776	0.1927
0.8935	2.1231	6000	1.1617	0.1891
2.5386	2.1939	6200	1.1686	0.1823
0.4705	2.2647	6400	1.1259	0.1843
1.2777	2.3355	6600	1.1228	0.1882
0.6823	2.4062	6800	1.1035	0.1787
1.2498	2.4770	7000	1.0976	0.1814
0.4579	2.5478	7200	1.0839	0.1721
0.9005	2.6185	7400	1.0695	0.1817
0.8045	2.6893	7600	1.0513	0.1745
1.3044	2.7601	7800	1.0336	0.1814
1.0797	2.8309	8000	1.0275	0.1846
0.3826	2.9016	8200	1.0178	0.1817
0.6175	2.9724	8400	1.0124	0.1764

Framework versions

• Transformers 4.40.0
• Pytorch 2.3.0+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1