layoutlmv3-finetuned-wildreceipt

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

• Loss: 0.3111
• Precision: 0.8749
• Recall: 0.8785
• F1: 0.8767
• Accuracy: 0.9253

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 4000

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.32	100	1.3060	0.6792	0.3615	0.4718	0.6966
No log	0.63	200	0.8842	0.6524	0.5193	0.5783	0.7737
No log	0.95	300	0.6795	0.7338	0.6772	0.7044	0.8336
No log	1.26	400	0.5604	0.7719	0.7390	0.7551	0.8629
1.0319	1.58	500	0.4862	0.7819	0.7618	0.7717	0.8730
1.0319	1.89	600	0.4365	0.7852	0.7807	0.7829	0.8795
1.0319	2.21	700	0.4182	0.8162	0.8016	0.8088	0.8897
1.0319	2.52	800	0.3886	0.8126	0.8196	0.8161	0.8936
1.0319	2.84	900	0.3637	0.8260	0.8347	0.8303	0.9004
0.4162	3.15	1000	0.3482	0.8532	0.8243	0.8385	0.9062
0.4162	3.47	1100	0.3474	0.8573	0.8248	0.8407	0.9042
0.4162	3.79	1200	0.3325	0.8408	0.8435	0.8421	0.9086
0.4162	4.1	1300	0.3262	0.8468	0.8467	0.8468	0.9095
0.4162	4.42	1400	0.3237	0.8511	0.8442	0.8477	0.9100
0.2764	4.73	1500	0.3156	0.8563	0.8456	0.8509	0.9122
0.2764	5.05	1600	0.3032	0.8558	0.8566	0.8562	0.9153
0.2764	5.36	1700	0.3120	0.8604	0.8457	0.8530	0.9142
0.2764	5.68	1800	0.2976	0.8608	0.8592	0.8600	0.9178
0.2764	5.99	1900	0.3056	0.8551	0.8676	0.8613	0.9171
0.212	6.31	2000	0.3191	0.8528	0.8599	0.8563	0.9147
0.212	6.62	2100	0.3051	0.8653	0.8635	0.8644	0.9186
0.212	6.94	2200	0.3022	0.8681	0.8632	0.8657	0.9208
0.212	7.26	2300	0.3101	0.8605	0.8643	0.8624	0.9178
0.212	7.57	2400	0.3100	0.8553	0.8693	0.8622	0.9163
0.1725	7.89	2500	0.3012	0.8685	0.8723	0.8704	0.9221
0.1725	8.2	2600	0.3135	0.8627	0.8756	0.8691	0.9187
0.1725	8.52	2700	0.3115	0.8768	0.8671	0.8719	0.9229
0.1725	8.83	2800	0.3044	0.8757	0.8708	0.8732	0.9231
0.1725	9.15	2900	0.3042	0.8698	0.8658	0.8678	0.9212
0.142	9.46	3000	0.3095	0.8677	0.8702	0.8690	0.9207
0.142	9.78	3100	0.3119	0.8686	0.8762	0.8724	0.9229
0.142	10.09	3200	0.3078	0.8713	0.8774	0.8743	0.9238
0.142	10.41	3300	0.3123	0.8711	0.8753	0.8732	0.9238
0.142	10.73	3400	0.3098	0.8688	0.8774	0.8731	0.9232
0.1238	11.04	3500	0.3120	0.8737	0.8770	0.8754	0.9247
0.1238	11.36	3600	0.3124	0.8760	0.8768	0.8764	0.9251
0.1238	11.67	3700	0.3101	0.8770	0.8759	0.8764	0.9254
0.1238	11.99	3800	0.3103	0.8767	0.8774	0.8770	0.9255
0.1238	12.3	3900	0.3122	0.8740	0.8788	0.8764	0.9251
0.1096	12.62	4000	0.3111	0.8749	0.8785	0.8767	0.9253

Framework versions

• Transformers 4.23.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.5.1
• Tokenizers 0.13.0