layoutlmv3-base-cord2

This model is a fine-tuned version of microsoft/layoutlmv3-base on the mp-02/cord dataset. It achieves the following results on the evaluation set:

• Loss: 0.1856
• Precision: 0.9467
• Recall: 0.9614
• F1: 0.9540
• Accuracy: 0.9611

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: 16
• eval_batch_size: 16
• 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: linear
• training_steps: 3000

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	100	1.2612	0.6788	0.7629	0.7184	0.7685
No log	2.0	200	0.5621	0.8674	0.8802	0.8738	0.8916
No log	3.0	300	0.3639	0.8846	0.9114	0.8978	0.9186
No log	4.0	400	0.3197	0.9153	0.9393	0.9271	0.9410
0.8719	5.0	500	0.2304	0.9357	0.9549	0.9452	0.9543
0.8719	6.0	600	0.2069	0.9389	0.9573	0.9480	0.9556
0.8719	7.0	700	0.2081	0.9459	0.9606	0.9532	0.9593
0.8719	8.0	800	0.1901	0.9532	0.9688	0.9609	0.9666
0.8719	9.0	900	0.1559	0.9515	0.9647	0.9580	0.9671
0.136	10.0	1000	0.1856	0.9467	0.9614	0.9540	0.9611
0.136	11.0	1100	0.2020	0.9537	0.9631	0.9584	0.9629
0.136	12.0	1200	0.1908	0.9552	0.9631	0.9592	0.9620

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3