invoice_extraction_donut_fromv0_f21_ep20_0724

This model is a fine-tuned version of naver-clova-ix/donut-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1221
• Char Accuracy: 0.6385

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: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Char Accuracy
0.5942	1.0	2991	0.2929	0.4026
0.2355	2.0	5982	0.1610	0.4363
0.2194	3.0	8973	0.0913	0.2990
0.0696	4.0	11964	0.0879	0.3076
0.0896	5.0	14955	0.0761	0.4217
0.0355	6.0	17946	0.0893	0.6547
0.0321	7.0	20937	0.0761	0.5625
0.0262	8.0	23928	0.0877	0.5487
0.0284	9.0	26919	0.0890	0.7065
0.0295	10.0	29910	0.0823	0.6354
0.0383	11.0	32901	0.0893	0.6316
0.0203	12.0	35892	0.0957	0.5729
0.0003	13.0	38883	0.0929	0.5490
0.0056	14.0	41874	0.1008	0.5576
0.0037	15.0	44865	0.1109	0.6604
0.0203	16.0	47856	0.1168	0.6210
0.0019	17.0	50847	0.1209	0.6400
0.0051	18.0	53838	0.1223	0.6334
0.0	19.0	56829	0.1216	0.6325
0.0053	20.0	59820	0.1221	0.6385

Framework versions

• Transformers 4.53.2
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1