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: 0.3129
• Precision: 0.8780
• Recall: 0.8870
• F1: 0.8825
• Accuracy: 0.9265

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	1.2240	0.6077	0.3766	0.4650	0.7011
No log	0.63	200	0.8417	0.6440	0.5089	0.5685	0.7743
No log	0.95	300	0.6466	0.7243	0.6583	0.6897	0.8311
No log	1.26	400	0.5516	0.7533	0.7158	0.7341	0.8537
0.9961	1.58	500	0.4845	0.7835	0.7557	0.7693	0.8699
0.9961	1.89	600	0.4506	0.7809	0.7930	0.7869	0.8770
0.9961	2.21	700	0.4230	0.8101	0.8107	0.8104	0.8886
0.9961	2.52	800	0.3797	0.8211	0.8296	0.8253	0.8983
0.9961	2.84	900	0.3576	0.8289	0.8411	0.8349	0.9016
0.4076	3.15	1000	0.3430	0.8394	0.8371	0.8382	0.9055
0.4076	3.47	1100	0.3354	0.8531	0.8405	0.8467	0.9071
0.4076	3.79	1200	0.3331	0.8371	0.8504	0.8437	0.9076
0.4076	4.1	1300	0.3184	0.8445	0.8609	0.8526	0.9118
0.4076	4.42	1400	0.3087	0.8617	0.8580	0.8598	0.9150
0.2673	4.73	1500	0.3013	0.8613	0.8657	0.8635	0.9177
0.2673	5.05	1600	0.2971	0.8630	0.8689	0.8659	0.9181
0.2673	5.36	1700	0.3075	0.8675	0.8639	0.8657	0.9177
0.2673	5.68	1800	0.2989	0.8551	0.8764	0.8656	0.9193
0.2673	5.99	1900	0.3011	0.8572	0.8762	0.8666	0.9194
0.2026	6.31	2000	0.3107	0.8595	0.8722	0.8658	0.9181
0.2026	6.62	2100	0.3050	0.8678	0.8800	0.8739	0.9220
0.2026	6.94	2200	0.2971	0.8722	0.8789	0.8755	0.9237
0.2026	7.26	2300	0.3057	0.8666	0.8785	0.8725	0.9209
0.2026	7.57	2400	0.3172	0.8593	0.8773	0.8682	0.9184
0.1647	7.89	2500	0.3018	0.8695	0.8823	0.8759	0.9228
0.1647	8.2	2600	0.3001	0.8760	0.8795	0.8777	0.9256
0.1647	8.52	2700	0.3068	0.8758	0.8745	0.8752	0.9235
0.1647	8.83	2800	0.3007	0.8779	0.8779	0.8779	0.9248
0.1647	9.15	2900	0.3063	0.8740	0.8763	0.8751	0.9228
0.1342	9.46	3000	0.3096	0.8675	0.8834	0.8754	0.9235
0.1342	9.78	3100	0.3052	0.8736	0.8848	0.8792	0.9249
0.1342	10.09	3200	0.3120	0.8727	0.8885	0.8805	0.9252
0.1342	10.41	3300	0.3146	0.8718	0.8843	0.8780	0.9243
0.1342	10.73	3400	0.3124	0.8720	0.8880	0.8799	0.9253
0.117	11.04	3500	0.3088	0.8761	0.8817	0.8789	0.9252
0.117	11.36	3600	0.3082	0.8782	0.8834	0.8808	0.9257
0.117	11.67	3700	0.3129	0.8767	0.8847	0.8807	0.9256
0.117	11.99	3800	0.3116	0.8792	0.8847	0.8820	0.9265
0.117	12.3	3900	0.3142	0.8768	0.8874	0.8821	0.9261
0.1022	12.62	4000	0.3129	0.8780	0.8870	0.8825	0.9265

Framework versions

• Transformers 4.22.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1