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: 1.7699
• Answer: {'precision': 0.8906439854191981, 'recall': 0.8971848225214198, 'f1': 0.8939024390243904, 'number': 817}
• Header: {'precision': 0.6274509803921569, 'recall': 0.5378151260504201, 'f1': 0.579185520361991, 'number': 119}
• Question: {'precision': 0.8778359511343804, 'recall': 0.9340761374187558, 'f1': 0.9050832208726944, 'number': 1077}
• Overall Precision: 0.8706
• Overall Recall: 0.8957
• Overall F1: 0.8830
• Overall Accuracy: 0.7973

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
0.4312	10.53	200	0.9853	{'precision': 0.8581818181818182, 'recall': 0.8665850673194615, 'f1': 0.8623629719853837, 'number': 817}	{'precision': 0.5625, 'recall': 0.5294117647058824, 'f1': 0.5454545454545455, 'number': 119}	{'precision': 0.8788706739526412, 'recall': 0.8960074280408542, 'f1': 0.8873563218390804, 'number': 1077}	0.8531	0.8624	0.8577	0.8172
0.0478	21.05	400	1.2825	{'precision': 0.8571428571428571, 'recall': 0.9033047735618115, 'f1': 0.8796185935637664, 'number': 817}	{'precision': 0.5136986301369864, 'recall': 0.6302521008403361, 'f1': 0.5660377358490567, 'number': 119}	{'precision': 0.8739650413983441, 'recall': 0.8820798514391829, 'f1': 0.878003696857671, 'number': 1077}	0.8419	0.8758	0.8585	0.8026
0.0127	31.58	600	1.4791	{'precision': 0.8568075117370892, 'recall': 0.8935128518971848, 'f1': 0.8747753145596165, 'number': 817}	{'precision': 0.5779816513761468, 'recall': 0.5294117647058824, 'f1': 0.5526315789473684, 'number': 119}	{'precision': 0.8909426987060998, 'recall': 0.8950789229340761, 'f1': 0.8930060213061601, 'number': 1077}	0.8600	0.8728	0.8664	0.7957
0.0073	42.11	800	1.3846	{'precision': 0.8853046594982079, 'recall': 0.9069767441860465, 'f1': 0.8960096735187424, 'number': 817}	{'precision': 0.5333333333333333, 'recall': 0.6050420168067226, 'f1': 0.5669291338582677, 'number': 119}	{'precision': 0.8932584269662921, 'recall': 0.8857938718662952, 'f1': 0.8895104895104896, 'number': 1077}	0.8662	0.8778	0.8719	0.8142
0.0023	52.63	1000	1.5955	{'precision': 0.8430034129692833, 'recall': 0.9069767441860465, 'f1': 0.8738207547169811, 'number': 817}	{'precision': 0.6190476190476191, 'recall': 0.5462184873949579, 'f1': 0.5803571428571429, 'number': 119}	{'precision': 0.8935574229691877, 'recall': 0.8885793871866295, 'f1': 0.8910614525139665, 'number': 1077}	0.8579	0.8758	0.8668	0.7992
0.0023	63.16	1200	1.6214	{'precision': 0.8955773955773956, 'recall': 0.8922888616891065, 'f1': 0.8939301042305334, 'number': 817}	{'precision': 0.5882352941176471, 'recall': 0.5882352941176471, 'f1': 0.5882352941176471, 'number': 119}	{'precision': 0.8841354723707665, 'recall': 0.9210770659238626, 'f1': 0.9022282855843565, 'number': 1077}	0.8715	0.8897	0.8805	0.8057
0.0016	73.68	1400	1.8002	{'precision': 0.8732394366197183, 'recall': 0.9106487148102815, 'f1': 0.8915518274415818, 'number': 817}	{'precision': 0.5765765765765766, 'recall': 0.5378151260504201, 'f1': 0.5565217391304348, 'number': 119}	{'precision': 0.8892921960072595, 'recall': 0.9099350046425255, 'f1': 0.8994951812758146, 'number': 1077}	0.8659	0.8882	0.8769	0.7860
0.0013	84.21	1600	1.7699	{'precision': 0.8906439854191981, 'recall': 0.8971848225214198, 'f1': 0.8939024390243904, 'number': 817}	{'precision': 0.6274509803921569, 'recall': 0.5378151260504201, 'f1': 0.579185520361991, 'number': 119}	{'precision': 0.8778359511343804, 'recall': 0.9340761374187558, 'f1': 0.9050832208726944, 'number': 1077}	0.8706	0.8957	0.8830	0.7973
0.0008	94.74	1800	1.7824	{'precision': 0.8733572281959379, 'recall': 0.8947368421052632, 'f1': 0.8839177750906893, 'number': 817}	{'precision': 0.616822429906542, 'recall': 0.5546218487394958, 'f1': 0.5840707964601769, 'number': 119}	{'precision': 0.8901996370235935, 'recall': 0.9108635097493036, 'f1': 0.9004130335016063, 'number': 1077}	0.8690	0.8833	0.8761	0.8019
0.0005	105.26	2000	1.7894	{'precision': 0.872791519434629, 'recall': 0.9069767441860465, 'f1': 0.8895558223289316, 'number': 817}	{'precision': 0.6036036036036037, 'recall': 0.5630252100840336, 'f1': 0.582608695652174, 'number': 119}	{'precision': 0.8931506849315068, 'recall': 0.9080779944289693, 'f1': 0.9005524861878452, 'number': 1077}	0.8691	0.8872	0.8781	0.7940
0.0002	115.79	2200	1.8409	{'precision': 0.8665893271461717, 'recall': 0.9143206854345165, 'f1': 0.8898153662894581, 'number': 817}	{'precision': 0.6296296296296297, 'recall': 0.5714285714285714, 'f1': 0.5991189427312775, 'number': 119}	{'precision': 0.8978644382544104, 'recall': 0.8978644382544104, 'f1': 0.8978644382544104, 'number': 1077}	0.8705	0.8852	0.8778	0.7982
0.0002	126.32	2400	1.8311	{'precision': 0.8709302325581395, 'recall': 0.9167686658506732, 'f1': 0.8932617769827073, 'number': 817}	{'precision': 0.6018518518518519, 'recall': 0.5462184873949579, 'f1': 0.5726872246696034, 'number': 119}	{'precision': 0.893953488372093, 'recall': 0.8922934076137419, 'f1': 0.8931226765799257, 'number': 1077}	0.8688	0.8818	0.8752	0.7988

Framework versions

• Transformers 4.25.1
• Pytorch 1.12.1
• Datasets 2.7.1
• Tokenizers 0.13.2