DonutInvoiceCzechV0123R

This model is a fine-tuned version of TomasFAV/DonutInvoiceCzechV012R on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2135
• Accuracy: 0.9240
• F1: 0.9094

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: 9e-05
• train_batch_size: 4
• eval_batch_size: 1
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.2043	1.0	46	0.1878	0.8989	0.8649
0.1079	2.0	92	0.1849	0.9008	0.8803
0.0613	3.0	138	0.1866	0.9030	0.8886
0.0396	4.0	184	0.2012	0.9185	0.8872
0.0263	5.0	230	0.2168	0.9105	0.8869
0.0213	6.0	276	0.2387	0.8936	0.8870
0.0183	7.0	322	0.2227	0.9122	0.8957
0.0199	8.0	368	0.2188	0.9170	0.8940
0.0167	9.0	414	0.2107	0.9188	0.9026
0.0225	10.0	460	0.2135	0.9240	0.9094
0.0089	11.0	506	0.2123	0.9203	0.8957
0.0097	12.0	552	0.2031	0.9196	0.9016
0.0049	13.0	598	0.2072	0.9177	0.8940
0.0024	14.0	644	0.2066	0.9180	0.9026
0.0006	15.0	690	0.2123	0.9185	0.9024
0.0005	16.0	736	0.2133	0.9174	0.8973
0.0019	17.0	782	0.2128	0.9172	0.8973
0.0012	18.0	828	0.2155	0.9215	0.9059
0.0013	19.0	874	0.2170	0.9138	0.9007
0.0029	20.0	920	0.2172	0.9138	0.9007

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2