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: 1.4588
• Answer: {'precision': 0.8786057692307693, 'recall': 0.8947368421052632, 'f1': 0.88659793814433, 'number': 817}
• Header: {'precision': 0.6442307692307693, 'recall': 0.5630252100840336, 'f1': 0.600896860986547, 'number': 119}
• Question: {'precision': 0.8854351687388987, 'recall': 0.9257195914577531, 'f1': 0.9051293690422152, 'number': 1077}
• Overall Precision: 0.8705
• Overall Recall: 0.8917
• Overall F1: 0.8810
• Overall Accuracy: 0.8222

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
0.4177	10.53	200	0.9741	{'precision': 0.834106728538283, 'recall': 0.8800489596083231, 'f1': 0.8564621798689696, 'number': 817}	{'precision': 0.6363636363636364, 'recall': 0.4117647058823529, 'f1': 0.5, 'number': 119}	{'precision': 0.8831985624438454, 'recall': 0.9127205199628597, 'f1': 0.897716894977169, 'number': 1077}	0.8533	0.8698	0.8615	0.8139
0.0528	21.05	400	1.2793	{'precision': 0.8391845979614949, 'recall': 0.9069767441860465, 'f1': 0.871764705882353, 'number': 817}	{'precision': 0.5480769230769231, 'recall': 0.4789915966386555, 'f1': 0.5112107623318385, 'number': 119}	{'precision': 0.878645343367827, 'recall': 0.8672237697307336, 'f1': 0.8728971962616822, 'number': 1077}	0.8449	0.8604	0.8526	0.8057
0.0154	31.58	600	1.3635	{'precision': 0.8705463182897862, 'recall': 0.8971848225214198, 'f1': 0.8836648583484027, 'number': 817}	{'precision': 0.6428571428571429, 'recall': 0.5294117647058824, 'f1': 0.5806451612903226, 'number': 119}	{'precision': 0.8776041666666666, 'recall': 0.9387186629526463, 'f1': 0.9071332436069988, 'number': 1077}	0.8638	0.8977	0.8804	0.8164
0.0082	42.11	800	1.4185	{'precision': 0.8700361010830325, 'recall': 0.8849449204406364, 'f1': 0.8774271844660194, 'number': 817}	{'precision': 0.6428571428571429, 'recall': 0.6050420168067226, 'f1': 0.6233766233766234, 'number': 119}	{'precision': 0.8921124206708976, 'recall': 0.9136490250696379, 'f1': 0.9027522935779816, 'number': 1077}	0.8695	0.8838	0.8766	0.8212
0.0038	52.63	1000	1.4588	{'precision': 0.8786057692307693, 'recall': 0.8947368421052632, 'f1': 0.88659793814433, 'number': 817}	{'precision': 0.6442307692307693, 'recall': 0.5630252100840336, 'f1': 0.600896860986547, 'number': 119}	{'precision': 0.8854351687388987, 'recall': 0.9257195914577531, 'f1': 0.9051293690422152, 'number': 1077}	0.8705	0.8917	0.8810	0.8222
0.0026	63.16	1200	1.5730	{'precision': 0.8666666666666667, 'recall': 0.8910648714810282, 'f1': 0.8786964393482196, 'number': 817}	{'precision': 0.7073170731707317, 'recall': 0.48739495798319327, 'f1': 0.5771144278606964, 'number': 119}	{'precision': 0.8887884267631103, 'recall': 0.9127205199628597, 'f1': 0.9005955107650022, 'number': 1077}	0.8723	0.8788	0.8755	0.8139
0.0015	73.68	1400	1.6294	{'precision': 0.837471783295711, 'recall': 0.9082007343941249, 'f1': 0.8714034057545508, 'number': 817}	{'precision': 0.6530612244897959, 'recall': 0.5378151260504201, 'f1': 0.5898617511520737, 'number': 119}	{'precision': 0.9039179104477612, 'recall': 0.8997214484679665, 'f1': 0.9018147975802697, 'number': 1077}	0.8633	0.8818	0.8725	0.8173
0.001	84.21	1600	1.6406	{'precision': 0.8434684684684685, 'recall': 0.9167686658506732, 'f1': 0.8785923753665689, 'number': 817}	{'precision': 0.6260869565217392, 'recall': 0.6050420168067226, 'f1': 0.6153846153846154, 'number': 119}	{'precision': 0.9001865671641791, 'recall': 0.8960074280408542, 'f1': 0.8980921358771522, 'number': 1077}	0.8607	0.8872	0.8738	0.8140
0.0006	94.74	1800	1.6743	{'precision': 0.8525714285714285, 'recall': 0.9130966952264382, 'f1': 0.8817966903073285, 'number': 817}	{'precision': 0.6666666666666666, 'recall': 0.5042016806722689, 'f1': 0.5741626794258373, 'number': 119}	{'precision': 0.8982584784601283, 'recall': 0.9099350046425255, 'f1': 0.904059040590406, 'number': 1077}	0.8687	0.8872	0.8779	0.8082
0.0003	105.26	2000	1.7003	{'precision': 0.8696682464454977, 'recall': 0.8984088127294981, 'f1': 0.8838049367850691, 'number': 817}	{'precision': 0.6404494382022472, 'recall': 0.4789915966386555, 'f1': 0.548076923076923, 'number': 119}	{'precision': 0.8927272727272727, 'recall': 0.9117920148560817, 'f1': 0.9021589343132751, 'number': 1077}	0.8721	0.8808	0.8764	0.8110
0.0002	115.79	2200	1.7767	{'precision': 0.8564867967853043, 'recall': 0.9130966952264382, 'f1': 0.8838862559241707, 'number': 817}	{'precision': 0.64, 'recall': 0.5378151260504201, 'f1': 0.5844748858447488, 'number': 119}	{'precision': 0.9077212806026366, 'recall': 0.8950789229340761, 'f1': 0.9013557737260401, 'number': 1077}	0.8726	0.8813	0.8769	0.8004
0.0002	126.32	2400	1.7093	{'precision': 0.8546910755148741, 'recall': 0.9143206854345165, 'f1': 0.8835008870490834, 'number': 817}	{'precision': 0.6413043478260869, 'recall': 0.4957983193277311, 'f1': 0.5592417061611374, 'number': 119}	{'precision': 0.8956602031394275, 'recall': 0.9006499535747446, 'f1': 0.8981481481481481, 'number': 1077}	0.8668	0.8823	0.8744	0.8027

Framework versions

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