TrOCRTraining2

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: 0.4589
• Cer: 0.0115

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: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.7331	1.06	50	0.8204	0.1775
0.4908	2.13	100	0.5457	0.0229
0.4912	3.19	150	0.5845	0.0229
0.4713	4.26	200	0.5433	0.0137
0.4435	5.32	250	0.4988	0.0126
0.4152	6.38	300	0.5058	0.0137
0.3026	7.45	350	0.4947	0.0126
0.4133	8.51	400	0.4988	0.0115
0.4029	9.57	450	0.4906	0.0160
0.3439	10.64	500	0.4790	0.0160
0.3386	11.7	550	0.4661	0.0103
0.3511	12.77	600	0.4617	0.0115
0.374	13.83	650	0.4629	0.0149
0.3357	14.89	700	0.4589	0.0115

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 2.12.0
• Tokenizers 0.13.3