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: 1.5278
• Answer: {'precision': 0.8726415094339622, 'recall': 0.9057527539779682, 'f1': 0.8888888888888888, 'number': 817}
• Header: {'precision': 0.6701030927835051, 'recall': 0.5462184873949579, 'f1': 0.6018518518518517, 'number': 119}
• Question: {'precision': 0.9128440366972477, 'recall': 0.9238625812441968, 'f1': 0.9183202584217812, 'number': 1077}
• Overall Precision: 0.8845
• Overall Recall: 0.8942
• Overall F1: 0.8893
• Overall Accuracy: 0.8213

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
0.4258	10.53	200	0.9877	{'precision': 0.8354143019296254, 'recall': 0.9008567931456548, 'f1': 0.866902237926973, 'number': 817}	{'precision': 0.5491803278688525, 'recall': 0.5630252100840336, 'f1': 0.5560165975103735, 'number': 119}	{'precision': 0.8691756272401434, 'recall': 0.9006499535747446, 'f1': 0.8846329229366164, 'number': 1077}	0.8367	0.8808	0.8582	0.8030
0.0396	21.05	400	1.2891	{'precision': 0.8314855875831486, 'recall': 0.9179926560587516, 'f1': 0.8726003490401396, 'number': 817}	{'precision': 0.5289256198347108, 'recall': 0.5378151260504201, 'f1': 0.5333333333333334, 'number': 119}	{'precision': 0.9008579599618685, 'recall': 0.8774373259052924, 'f1': 0.8889934148635935, 'number': 1077}	0.8489	0.8738	0.8612	0.8071
0.0145	31.58	600	1.1878	{'precision': 0.8541666666666666, 'recall': 0.9033047735618115, 'f1': 0.8780487804878049, 'number': 817}	{'precision': 0.5943396226415094, 'recall': 0.5294117647058824, 'f1': 0.5599999999999999, 'number': 119}	{'precision': 0.8793565683646113, 'recall': 0.9136490250696379, 'f1': 0.896174863387978, 'number': 1077}	0.8545	0.8867	0.8703	0.8139
0.0093	42.11	800	1.3968	{'precision': 0.8727272727272727, 'recall': 0.8812729498164015, 'f1': 0.8769792935444579, 'number': 817}	{'precision': 0.5454545454545454, 'recall': 0.6050420168067226, 'f1': 0.5737051792828685, 'number': 119}	{'precision': 0.8951686417502279, 'recall': 0.9117920148560817, 'f1': 0.9034038638454462, 'number': 1077}	0.8637	0.8813	0.8724	0.8054
0.0042	52.63	1000	1.5509	{'precision': 0.8372093023255814, 'recall': 0.9253365973072215, 'f1': 0.8790697674418605, 'number': 817}	{'precision': 0.6304347826086957, 'recall': 0.48739495798319327, 'f1': 0.5497630331753555, 'number': 119}	{'precision': 0.9044048734770385, 'recall': 0.8960074280408542, 'f1': 0.9001865671641791, 'number': 1077}	0.8628	0.8838	0.8731	0.8044
0.0026	63.16	1200	1.5696	{'precision': 0.8618266978922716, 'recall': 0.9008567931456548, 'f1': 0.8809096349491322, 'number': 817}	{'precision': 0.6666666666666666, 'recall': 0.5210084033613446, 'f1': 0.5849056603773585, 'number': 119}	{'precision': 0.8935978358881875, 'recall': 0.9201485608170845, 'f1': 0.9066788655077767, 'number': 1077}	0.8701	0.8887	0.8793	0.8116
0.001	73.68	1400	1.7209	{'precision': 0.8396860986547086, 'recall': 0.9167686658506732, 'f1': 0.8765359859566998, 'number': 817}	{'precision': 0.6781609195402298, 'recall': 0.4957983193277311, 'f1': 0.5728155339805825, 'number': 119}	{'precision': 0.8969359331476323, 'recall': 0.8969359331476323, 'f1': 0.8969359331476322, 'number': 1077}	0.8628	0.8813	0.8720	0.7977
0.0011	84.21	1600	1.5329	{'precision': 0.8646188850967008, 'recall': 0.9302325581395349, 'f1': 0.8962264150943396, 'number': 817}	{'precision': 0.6666666666666666, 'recall': 0.5042016806722689, 'f1': 0.5741626794258373, 'number': 119}	{'precision': 0.9050691244239631, 'recall': 0.9117920148560817, 'f1': 0.9084181313598519, 'number': 1077}	0.8773	0.8952	0.8862	0.8267
0.0006	94.74	1800	1.5523	{'precision': 0.8748510131108462, 'recall': 0.8984088127294981, 'f1': 0.8864734299516908, 'number': 817}	{'precision': 0.5811965811965812, 'recall': 0.5714285714285714, 'f1': 0.576271186440678, 'number': 119}	{'precision': 0.9045412418906394, 'recall': 0.9062209842154132, 'f1': 0.9053803339517627, 'number': 1077}	0.8737	0.8833	0.8785	0.8196
0.0005	105.26	2000	1.5178	{'precision': 0.8758949880668258, 'recall': 0.8984088127294981, 'f1': 0.8870090634441088, 'number': 817}	{'precision': 0.6428571428571429, 'recall': 0.5294117647058824, 'f1': 0.5806451612903226, 'number': 119}	{'precision': 0.8995475113122172, 'recall': 0.9229340761374187, 'f1': 0.9110907424381303, 'number': 1077}	0.8775	0.8897	0.8836	0.8253
0.0004	115.79	2200	1.5493	{'precision': 0.8597701149425288, 'recall': 0.9155446756425949, 'f1': 0.8867812685240072, 'number': 817}	{'precision': 0.6631578947368421, 'recall': 0.5294117647058824, 'f1': 0.5887850467289719, 'number': 119}	{'precision': 0.9107635694572217, 'recall': 0.9192200557103064, 'f1': 0.9149722735674676, 'number': 1077}	0.8777	0.8947	0.8861	0.8217
0.0003	126.32	2400	1.5278	{'precision': 0.8726415094339622, 'recall': 0.9057527539779682, 'f1': 0.8888888888888888, 'number': 817}	{'precision': 0.6701030927835051, 'recall': 0.5462184873949579, 'f1': 0.6018518518518517, 'number': 119}	{'precision': 0.9128440366972477, 'recall': 0.9238625812441968, 'f1': 0.9183202584217812, 'number': 1077}	0.8845	0.8942	0.8893	0.8213

Framework versions

• Transformers 4.28.1
• Pytorch 1.13.1+cu117
• Datasets 2.11.0
• Tokenizers 0.13.3