lilt-en-funsd

This model is a fine-tuned version of SCUT-DLVCLab/lilt-roberta-en-base on the funsd-layoutlmv3 dataset. It achieves the following results on the evaluation set:

• Loss: 1.6502
• Answer: {'precision': 0.8637413394919169, 'recall': 0.9155446756425949, 'f1': 0.888888888888889, 'number': 817}
• Header: {'precision': 0.625, 'recall': 0.5042016806722689, 'f1': 0.5581395348837209, 'number': 119}
• Question: {'precision': 0.8934280639431617, 'recall': 0.9340761374187558, 'f1': 0.9133000453926464, 'number': 1077}
• Overall Precision: 0.8688
• Overall Recall: 0.9011
• Overall F1: 0.8847
• Overall Accuracy: 0.8015

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: 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: 2500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Answer	Header	Question	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
0.4362	10.53	200	0.9773	{'precision': 0.8193832599118943, 'recall': 0.9106487148102815, 'f1': 0.862608695652174, 'number': 817}	{'precision': 0.6666666666666666, 'recall': 0.3865546218487395, 'f1': 0.4893617021276595, 'number': 119}	{'precision': 0.8725490196078431, 'recall': 0.9090064995357474, 'f1': 0.8904047294224647, 'number': 1077}	0.8428	0.8788	0.8604	0.7932
0.0418	21.05	400	1.4204	{'precision': 0.8056460369163952, 'recall': 0.9082007343941249, 'f1': 0.85385500575374, 'number': 817}	{'precision': 0.5684210526315789, 'recall': 0.453781512605042, 'f1': 0.5046728971962617, 'number': 119}	{'precision': 0.8854262144821264, 'recall': 0.8969359331476323, 'f1': 0.8911439114391144, 'number': 1077}	0.8363	0.8753	0.8553	0.7870
0.0118	31.58	600	1.5084	{'precision': 0.8661137440758294, 'recall': 0.8947368421052632, 'f1': 0.8801926550270921, 'number': 817}	{'precision': 0.5575221238938053, 'recall': 0.5294117647058824, 'f1': 0.543103448275862, 'number': 119}	{'precision': 0.8864864864864865, 'recall': 0.9136490250696379, 'f1': 0.8998628257887517, 'number': 1077}	0.8602	0.8833	0.8716	0.7938
0.0116	42.11	800	1.4934	{'precision': 0.8497109826589595, 'recall': 0.8996328029375765, 'f1': 0.8739595719381689, 'number': 817}	{'precision': 0.6068376068376068, 'recall': 0.5966386554621849, 'f1': 0.6016949152542374, 'number': 119}	{'precision': 0.8835489833641405, 'recall': 0.8876508820798514, 'f1': 0.8855951829550718, 'number': 1077}	0.8537	0.8753	0.8644	0.7963
0.0046	52.63	1000	1.6502	{'precision': 0.8637413394919169, 'recall': 0.9155446756425949, 'f1': 0.888888888888889, 'number': 817}	{'precision': 0.625, 'recall': 0.5042016806722689, 'f1': 0.5581395348837209, 'number': 119}	{'precision': 0.8934280639431617, 'recall': 0.9340761374187558, 'f1': 0.9133000453926464, 'number': 1077}	0.8688	0.9011	0.8847	0.8015
0.0025	63.16	1200	1.6009	{'precision': 0.8503480278422274, 'recall': 0.8971848225214198, 'f1': 0.8731387730792138, 'number': 817}	{'precision': 0.651685393258427, 'recall': 0.48739495798319327, 'f1': 0.5576923076923077, 'number': 119}	{'precision': 0.8716392020815265, 'recall': 0.9331476323119777, 'f1': 0.9013452914798208, 'number': 1077}	0.8536	0.8922	0.8725	0.8073
0.0016	73.68	1400	1.6601	{'precision': 0.8872727272727273, 'recall': 0.8959608323133414, 'f1': 0.8915956151035324, 'number': 817}	{'precision': 0.67, 'recall': 0.5630252100840336, 'f1': 0.6118721461187214, 'number': 119}	{'precision': 0.8820375335120644, 'recall': 0.9164345403899722, 'f1': 0.8989071038251366, 'number': 1077}	0.8738	0.8872	0.8805	0.7977
0.0006	84.21	1600	1.6735	{'precision': 0.8774038461538461, 'recall': 0.8935128518971848, 'f1': 0.8853850818677986, 'number': 817}	{'precision': 0.6636363636363637, 'recall': 0.6134453781512605, 'f1': 0.6375545851528385, 'number': 119}	{'precision': 0.8782452999104745, 'recall': 0.9108635097493036, 'f1': 0.8942570647219691, 'number': 1077}	0.8664	0.8862	0.8762	0.7997
0.0006	94.74	1800	1.6672	{'precision': 0.8755980861244019, 'recall': 0.8959608323133414, 'f1': 0.8856624319419237, 'number': 817}	{'precision': 0.6545454545454545, 'recall': 0.6050420168067226, 'f1': 0.62882096069869, 'number': 119}	{'precision': 0.8800705467372134, 'recall': 0.9266480965645311, 'f1': 0.902758932609679, 'number': 1077}	0.8663	0.8952	0.8805	0.8010
0.0004	105.26	2000	1.6652	{'precision': 0.8880866425992779, 'recall': 0.9033047735618115, 'f1': 0.895631067961165, 'number': 817}	{'precision': 0.6086956521739131, 'recall': 0.5882352941176471, 'f1': 0.5982905982905983, 'number': 119}	{'precision': 0.8776785714285714, 'recall': 0.9127205199628597, 'f1': 0.8948566226672735, 'number': 1077}	0.8669	0.8897	0.8782	0.8059
0.0004	115.79	2200	1.6698	{'precision': 0.8993865030674847, 'recall': 0.8971848225214198, 'f1': 0.8982843137254903, 'number': 817}	{'precision': 0.631578947368421, 'recall': 0.6050420168067226, 'f1': 0.6180257510729613, 'number': 119}	{'precision': 0.8808243727598566, 'recall': 0.9127205199628597, 'f1': 0.8964888280893752, 'number': 1077}	0.8743	0.8882	0.8812	0.8096
0.0002	126.32	2400	1.7190	{'precision': 0.8888888888888888, 'recall': 0.9008567931456548, 'f1': 0.8948328267477204, 'number': 817}	{'precision': 0.6542056074766355, 'recall': 0.5882352941176471, 'f1': 0.6194690265486726, 'number': 119}	{'precision': 0.8815672306322351, 'recall': 0.9192200557103064, 'f1': 0.9, 'number': 1077}	0.8727	0.8922	0.8823	0.8045

Framework versions

• Transformers 4.27.3
• Pytorch 1.13.1+cu116
• Datasets 2.10.1
• Tokenizers 0.13.2