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: 0.3154
• Precision: 0.8693
• Recall: 0.8753
• F1: 0.8723
• Accuracy: 0.9240

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	1.3618	0.6375	0.3049	0.4125	0.6708
No log	0.63	200	0.9129	0.6662	0.4897	0.5645	0.7631
No log	0.95	300	0.6800	0.7273	0.6375	0.6795	0.8274
No log	1.26	400	0.5733	0.7579	0.6926	0.7238	0.8501
1.0638	1.58	500	0.5015	0.7854	0.7383	0.7611	0.8667
1.0638	1.89	600	0.4501	0.7916	0.7680	0.7796	0.8770
1.0638	2.21	700	0.4145	0.8177	0.8053	0.8114	0.8917
1.0638	2.52	800	0.3835	0.8214	0.8210	0.8212	0.8961
1.0638	2.84	900	0.3666	0.8251	0.8338	0.8294	0.9009
0.423	3.15	1000	0.3524	0.8485	0.8217	0.8349	0.9026
0.423	3.47	1100	0.3451	0.8430	0.8327	0.8378	0.9060
0.423	3.79	1200	0.3348	0.8347	0.8504	0.8425	0.9092
0.423	4.1	1300	0.3331	0.8368	0.8448	0.8408	0.9079
0.423	4.42	1400	0.3163	0.8503	0.8519	0.8511	0.9138
0.2822	4.73	1500	0.3168	0.8531	0.8504	0.8518	0.9133
0.2822	5.05	1600	0.3013	0.8629	0.8577	0.8603	0.9183
0.2822	5.36	1700	0.3146	0.8619	0.8528	0.8573	0.9160
0.2822	5.68	1800	0.3121	0.8474	0.8638	0.8555	0.9159
0.2822	5.99	1900	0.3054	0.8537	0.8667	0.8601	0.9166
0.2176	6.31	2000	0.3127	0.8556	0.8592	0.8574	0.9167
0.2176	6.62	2100	0.3072	0.8568	0.8667	0.8617	0.9194
0.2176	6.94	2200	0.2989	0.8617	0.8660	0.8638	0.9209
0.2176	7.26	2300	0.2997	0.8616	0.8682	0.8649	0.9199
0.2176	7.57	2400	0.3100	0.8538	0.8689	0.8613	0.9191
0.1777	7.89	2500	0.3022	0.8649	0.8695	0.8672	0.9228
0.1777	8.2	2600	0.2990	0.8631	0.8740	0.8685	0.9224
0.1777	8.52	2700	0.3072	0.8669	0.8697	0.8683	0.9228
0.1777	8.83	2800	0.3038	0.8689	0.8720	0.8705	0.9238
0.1777	9.15	2900	0.3138	0.8726	0.8673	0.8700	0.9216
0.1434	9.46	3000	0.3150	0.8610	0.8740	0.8674	0.9221
0.1434	9.78	3100	0.3055	0.8674	0.8742	0.8708	0.9239
0.1434	10.09	3200	0.3182	0.8618	0.8724	0.8671	0.9215
0.1434	10.41	3300	0.3175	0.8633	0.8727	0.8680	0.9223
0.1434	10.73	3400	0.3146	0.8685	0.8717	0.8701	0.9234
0.1282	11.04	3500	0.3136	0.8671	0.8757	0.8714	0.9233
0.1282	11.36	3600	0.3186	0.8679	0.8734	0.8706	0.9225
0.1282	11.67	3700	0.3147	0.8701	0.8745	0.8723	0.9238
0.1282	11.99	3800	0.3159	0.8705	0.8759	0.8732	0.9244
0.1282	12.3	3900	0.3147	0.8699	0.8748	0.8723	0.9246
0.1121	12.62	4000	0.3154	0.8693	0.8753	0.8723	0.9240

Framework versions

• Transformers 4.22.0.dev0
• Pytorch 1.12.0+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1