layoutlmv3-invoice

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

• Loss: 0.1889
• Precision: 0.9698
• Recall: 0.9592
• F1: 0.9645
• Accuracy: 0.9708

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: 1e-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
• training_steps: 3500

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	5.32	250	0.3826	0.9471	0.9278	0.9374	0.9465
0.8433	10.64	500	0.1720	0.9697	0.9560	0.9628	0.9684
0.8433	15.96	750	0.1631	0.9714	0.9608	0.9661	0.9684
0.0347	21.28	1000	0.1548	0.9746	0.9639	0.9692	0.9733
0.0347	26.6	1250	0.1700	0.9698	0.9576	0.9637	0.9672
0.0116	31.91	1500	0.1812	0.9667	0.9576	0.9621	0.9648
0.0116	37.23	1750	0.1513	0.9683	0.9592	0.9637	0.9721
0.0066	42.55	2000	0.1555	0.9730	0.9623	0.9676	0.9757
0.0066	47.87	2250	0.1729	0.9714	0.9592	0.9652	0.9708
0.0048	53.19	2500	0.1854	0.9761	0.9623	0.9692	0.9721
0.0048	58.51	2750	0.1863	0.9714	0.9592	0.9652	0.9696
0.0037	63.83	3000	0.1813	0.9761	0.9623	0.9692	0.9733
0.0037	69.15	3250	0.1903	0.9698	0.9592	0.9645	0.9708
0.0034	74.47	3500	0.1889	0.9698	0.9592	0.9645	0.9708

Framework versions

• Transformers 4.36.0.dev0
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0