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: 1.6055
• Answer: {'precision': 0.8728323699421965, 'recall': 0.9241126070991432, 'f1': 0.8977407847800237, 'number': 817}
• Header: {'precision': 0.6, 'recall': 0.5042016806722689, 'f1': 0.547945205479452, 'number': 119}
• Question: {'precision': 0.9, 'recall': 0.9108635097493036, 'f1': 0.9053991693585604, 'number': 1077}
• Overall Precision: 0.8740
• Overall Recall: 0.8922
• Overall F1: 0.8830
• Overall Accuracy: 0.8022

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
0.4096	10.53	200	0.9947	{'precision': 0.8293515358361775, 'recall': 0.8922888616891065, 'f1': 0.8596698113207547, 'number': 817}	{'precision': 0.5833333333333334, 'recall': 0.5294117647058824, 'f1': 0.5550660792951542, 'number': 119}	{'precision': 0.8858195211786372, 'recall': 0.89322191272052, 'f1': 0.8895053166897827, 'number': 1077}	0.8461	0.8713	0.8585	0.8133
0.0451	21.05	400	1.4850	{'precision': 0.8394495412844036, 'recall': 0.8959608323133414, 'f1': 0.866785079928952, 'number': 817}	{'precision': 0.6333333333333333, 'recall': 0.4789915966386555, 'f1': 0.5454545454545454, 'number': 119}	{'precision': 0.9026974951830443, 'recall': 0.8700092850510678, 'f1': 0.8860520094562647, 'number': 1077}	0.863	0.8574	0.8602	0.7910
0.0147	31.58	600	1.5603	{'precision': 0.8478260869565217, 'recall': 0.9069767441860465, 'f1': 0.8764044943820224, 'number': 817}	{'precision': 0.6336633663366337, 'recall': 0.5378151260504201, 'f1': 0.5818181818181819, 'number': 119}	{'precision': 0.8845807033363391, 'recall': 0.9108635097493036, 'f1': 0.8975297346752059, 'number': 1077}	0.8570	0.8872	0.8719	0.7948
0.0077	42.11	800	1.7433	{'precision': 0.8344444444444444, 'recall': 0.9192166462668299, 'f1': 0.8747815958066395, 'number': 817}	{'precision': 0.5596330275229358, 'recall': 0.5126050420168067, 'f1': 0.5350877192982455, 'number': 119}	{'precision': 0.9031954887218046, 'recall': 0.8922934076137419, 'f1': 0.897711349836525, 'number': 1077}	0.8553	0.8808	0.8678	0.7910
0.004	52.63	1000	1.6055	{'precision': 0.8728323699421965, 'recall': 0.9241126070991432, 'f1': 0.8977407847800237, 'number': 817}	{'precision': 0.6, 'recall': 0.5042016806722689, 'f1': 0.547945205479452, 'number': 119}	{'precision': 0.9, 'recall': 0.9108635097493036, 'f1': 0.9053991693585604, 'number': 1077}	0.8740	0.8922	0.8830	0.8022
0.0021	63.16	1200	1.7317	{'precision': 0.8688524590163934, 'recall': 0.9082007343941249, 'f1': 0.8880909634949131, 'number': 817}	{'precision': 0.5904761904761905, 'recall': 0.5210084033613446, 'f1': 0.5535714285714286, 'number': 119}	{'precision': 0.8847184986595175, 'recall': 0.9192200557103064, 'f1': 0.9016393442622952, 'number': 1077}	0.8633	0.8912	0.8770	0.7979
0.0017	73.68	1400	1.7249	{'precision': 0.8705463182897862, 'recall': 0.8971848225214198, 'f1': 0.8836648583484027, 'number': 817}	{'precision': 0.5555555555555556, 'recall': 0.5042016806722689, 'f1': 0.5286343612334802, 'number': 119}	{'precision': 0.8818181818181818, 'recall': 0.9006499535747446, 'f1': 0.891134588883785, 'number': 1077}	0.86	0.8758	0.8678	0.8014
0.0012	84.21	1600	1.8716	{'precision': 0.8679678530424799, 'recall': 0.9253365973072215, 'f1': 0.8957345971563981, 'number': 817}	{'precision': 0.5652173913043478, 'recall': 0.5462184873949579, 'f1': 0.5555555555555555, 'number': 119}	{'precision': 0.898320895522388, 'recall': 0.8941504178272981, 'f1': 0.8962308050255934, 'number': 1077}	0.8669	0.8862	0.8764	0.7946
0.0006	94.74	1800	1.8381	{'precision': 0.8563218390804598, 'recall': 0.9118727050183598, 'f1': 0.8832246591582691, 'number': 817}	{'precision': 0.6055045871559633, 'recall': 0.5546218487394958, 'f1': 0.5789473684210525, 'number': 119}	{'precision': 0.8931226765799256, 'recall': 0.8922934076137419, 'f1': 0.8927078495123084, 'number': 1077}	0.8623	0.8803	0.8712	0.7954
0.0007	105.26	2000	1.7090	{'precision': 0.8822115384615384, 'recall': 0.8984088127294981, 'f1': 0.8902365069739235, 'number': 817}	{'precision': 0.576271186440678, 'recall': 0.5714285714285714, 'f1': 0.5738396624472574, 'number': 119}	{'precision': 0.8811881188118812, 'recall': 0.9090064995357474, 'f1': 0.8948811700182815, 'number': 1077}	0.8641	0.8847	0.8743	0.8122
0.0003	115.79	2200	1.7487	{'precision': 0.8730723606168446, 'recall': 0.9008567931456548, 'f1': 0.8867469879518072, 'number': 817}	{'precision': 0.5645161290322581, 'recall': 0.5882352941176471, 'f1': 0.5761316872427984, 'number': 119}	{'precision': 0.8936755270394133, 'recall': 0.9052924791086351, 'f1': 0.8994464944649446, 'number': 1077}	0.8654	0.8847	0.8750	0.8084
0.0002	126.32	2400	1.7644	{'precision': 0.8686046511627907, 'recall': 0.9143206854345165, 'f1': 0.8908765652951699, 'number': 817}	{'precision': 0.5932203389830508, 'recall': 0.5882352941176471, 'f1': 0.5907172995780592, 'number': 119}	{'precision': 0.8961397058823529, 'recall': 0.9052924791086351, 'f1': 0.9006928406466513, 'number': 1077}	0.8674	0.8902	0.8786	0.8091

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1+cu118
• Datasets 2.12.0
• Tokenizers 0.13.3