layoutlmv3-invoice-model

This model is a fine-tuned version of microsoft/layoutlmv3-base on the ls-generated4 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3694
• Precision: 0.9186
• Recall: 0.9375
• F1: 0.9279
• Accuracy: 0.9537

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 2000

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.85	100	0.7836	0.5238	0.5982	0.5585	0.7680
No log	1.69	200	0.4954	0.6888	0.7479	0.7172	0.8422
No log	2.54	300	0.3483	0.7807	0.8462	0.8121	0.9040
No log	3.39	400	0.3200	0.8113	0.8654	0.8375	0.9125
0.5923	4.24	500	0.2775	0.8593	0.8853	0.8721	0.9319
0.5923	5.08	600	0.2674	0.8700	0.9052	0.8872	0.9377
0.5923	5.93	700	0.2766	0.8739	0.9135	0.8932	0.9386
0.5923	6.78	800	0.2641	0.8879	0.9190	0.9031	0.9472
0.5923	7.63	900	0.2893	0.9094	0.9238	0.9165	0.9447
0.0802	8.47	1000	0.3369	0.9145	0.9258	0.9201	0.9465
0.0802	9.32	1100	0.3037	0.9043	0.9341	0.9189	0.9505
0.0802	10.17	1200	0.3510	0.9032	0.9231	0.9130	0.9472
0.0802	11.02	1300	0.3224	0.9138	0.9251	0.9195	0.9501
0.0802	11.86	1400	0.3873	0.9133	0.9265	0.9199	0.9456
0.0198	12.71	1500	0.3786	0.9120	0.9327	0.9222	0.9492
0.0198	13.56	1600	0.3807	0.9050	0.9293	0.9170	0.9469
0.0198	14.41	1700	0.3664	0.9088	0.9313	0.9199	0.9510
0.0198	15.25	1800	0.3582	0.9152	0.9341	0.9245	0.9521
0.0198	16.1	1900	0.3736	0.9198	0.9368	0.9282	0.9528
0.007	16.95	2000	0.3694	0.9186	0.9375	0.9279	0.9537

Framework versions

• Transformers 4.35.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1