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: 1.6250
• Answer: {'precision': 0.8670520231213873, 'recall': 0.9179926560587516, 'f1': 0.89179548156956, 'number': 817}
• Header: {'precision': 0.6796116504854369, 'recall': 0.5882352941176471, 'f1': 0.6306306306306307, 'number': 119}
• Question: {'precision': 0.902867715078631, 'recall': 0.9062209842154132, 'f1': 0.9045412418906396, 'number': 1077}
• Overall Precision: 0.8765
• Overall Recall: 0.8922
• Overall F1: 0.8843
• Overall Accuracy: 0.8191

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
0.4102	10.53	200	0.9780	{'precision': 0.794341675734494, 'recall': 0.8935128518971848, 'f1': 0.8410138248847926, 'number': 817}	{'precision': 0.6351351351351351, 'recall': 0.3949579831932773, 'f1': 0.48704663212435234, 'number': 119}	{'precision': 0.8618834080717489, 'recall': 0.8922934076137419, 'f1': 0.8768248175182481, 'number': 1077}	0.8245	0.8634	0.8435	0.8098
0.0415	21.05	400	1.2998	{'precision': 0.8573113207547169, 'recall': 0.8898408812729498, 'f1': 0.8732732732732732, 'number': 817}	{'precision': 0.5887850467289719, 'recall': 0.5294117647058824, 'f1': 0.5575221238938053, 'number': 119}	{'precision': 0.8603256212510711, 'recall': 0.9322191272051996, 'f1': 0.894830659536542, 'number': 1077}	0.8454	0.8912	0.8677	0.8044
0.0138	31.58	600	1.4296	{'precision': 0.8388952819332566, 'recall': 0.8922888616891065, 'f1': 0.8647686832740212, 'number': 817}	{'precision': 0.5185185185185185, 'recall': 0.7058823529411765, 'f1': 0.597864768683274, 'number': 119}	{'precision': 0.906158357771261, 'recall': 0.8607242339832869, 'f1': 0.8828571428571429, 'number': 1077}	0.8471	0.8644	0.8557	0.8033
0.0071	42.11	800	1.5437	{'precision': 0.8325991189427313, 'recall': 0.9253365973072215, 'f1': 0.8765217391304347, 'number': 817}	{'precision': 0.6593406593406593, 'recall': 0.5042016806722689, 'f1': 0.5714285714285715, 'number': 119}	{'precision': 0.8944392082940622, 'recall': 0.8811513463324049, 'f1': 0.8877455565949485, 'number': 1077}	0.8568	0.8768	0.8667	0.8003
0.0035	52.63	1000	1.6306	{'precision': 0.8327832783278328, 'recall': 0.9265605875152999, 'f1': 0.8771726535341833, 'number': 817}	{'precision': 0.6509433962264151, 'recall': 0.5798319327731093, 'f1': 0.6133333333333333, 'number': 119}	{'precision': 0.9034676663542643, 'recall': 0.8950789229340761, 'f1': 0.8992537313432836, 'number': 1077}	0.8598	0.8892	0.8742	0.7967
0.0022	63.16	1200	1.6872	{'precision': 0.8472063854047891, 'recall': 0.9094247246022031, 'f1': 0.8772136953955136, 'number': 817}	{'precision': 0.6363636363636364, 'recall': 0.5294117647058824, 'f1': 0.5779816513761468, 'number': 119}	{'precision': 0.9077212806026366, 'recall': 0.8950789229340761, 'f1': 0.9013557737260401, 'number': 1077}	0.8685	0.8793	0.8739	0.7997
0.0021	73.68	1400	1.6366	{'precision': 0.8106060606060606, 'recall': 0.9167686658506732, 'f1': 0.8604250430786904, 'number': 817}	{'precision': 0.5904761904761905, 'recall': 0.5210084033613446, 'f1': 0.5535714285714286, 'number': 119}	{'precision': 0.8941605839416058, 'recall': 0.9099350046425255, 'f1': 0.9019788311090657, 'number': 1077}	0.8428	0.8897	0.8656	0.8054
0.0011	84.21	1600	1.5864	{'precision': 0.8795180722891566, 'recall': 0.8935128518971848, 'f1': 0.8864602307225258, 'number': 817}	{'precision': 0.6481481481481481, 'recall': 0.5882352941176471, 'f1': 0.6167400881057269, 'number': 119}	{'precision': 0.8894783377541998, 'recall': 0.9340761374187558, 'f1': 0.911231884057971, 'number': 1077}	0.8729	0.8972	0.8849	0.8194
0.0005	94.74	1800	1.5746	{'precision': 0.8587699316628702, 'recall': 0.9228886168910648, 'f1': 0.8896755162241888, 'number': 817}	{'precision': 0.66, 'recall': 0.5546218487394958, 'f1': 0.6027397260273973, 'number': 119}	{'precision': 0.9055045871559633, 'recall': 0.9164345403899722, 'f1': 0.9109367789570835, 'number': 1077}	0.8738	0.8977	0.8856	0.8254
0.0004	105.26	2000	1.6031	{'precision': 0.8669778296382731, 'recall': 0.9094247246022031, 'f1': 0.8876941457586618, 'number': 817}	{'precision': 0.6173913043478261, 'recall': 0.5966386554621849, 'f1': 0.6068376068376068, 'number': 119}	{'precision': 0.904363974001857, 'recall': 0.904363974001857, 'f1': 0.904363974001857, 'number': 1077}	0.8726	0.8882	0.8804	0.8218
0.0003	115.79	2200	1.6122	{'precision': 0.8632183908045977, 'recall': 0.9192166462668299, 'f1': 0.890337877889745, 'number': 817}	{'precision': 0.6831683168316832, 'recall': 0.5798319327731093, 'f1': 0.6272727272727273, 'number': 119}	{'precision': 0.9016544117647058, 'recall': 0.9108635097493036, 'f1': 0.9062355658198614, 'number': 1077}	0.8747	0.8947	0.8846	0.8221
0.0002	126.32	2400	1.6250	{'precision': 0.8670520231213873, 'recall': 0.9179926560587516, 'f1': 0.89179548156956, 'number': 817}	{'precision': 0.6796116504854369, 'recall': 0.5882352941176471, 'f1': 0.6306306306306307, 'number': 119}	{'precision': 0.902867715078631, 'recall': 0.9062209842154132, 'f1': 0.9045412418906396, 'number': 1077}	0.8765	0.8922	0.8843	0.8191

Framework versions

• Transformers 4.28.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.13.3