LiLt-funsd-en

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: 1.7735
• Answer: {'precision': 0.8693115519253208, 'recall': 0.9118727050183598, 'f1': 0.8900836320191159, 'number': 817}
• Header: {'precision': 0.6630434782608695, 'recall': 0.5126050420168067, 'f1': 0.5781990521327014, 'number': 119}
• Question: {'precision': 0.8992673992673993, 'recall': 0.9117920148560817, 'f1': 0.905486399262333, 'number': 1077}
• Overall Precision: 0.8760
• Overall Recall: 0.8882
• Overall F1: 0.8821
• Overall Accuracy: 0.8027

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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Answer	Header	Question	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
0.4108	10.53	200	0.9495	{'precision': 0.7949260042283298, 'recall': 0.9204406364749081, 'f1': 0.8530913216108904, 'number': 817}	{'precision': 0.5338345864661654, 'recall': 0.5966386554621849, 'f1': 0.5634920634920635, 'number': 119}	{'precision': 0.8950914340712224, 'recall': 0.8635097493036211, 'f1': 0.8790170132325141, 'number': 1077}	0.8277	0.8708	0.8487	0.7845
0.0402	21.05	400	1.3124	{'precision': 0.8167597765363128, 'recall': 0.8947368421052632, 'f1': 0.8539719626168224, 'number': 817}	{'precision': 0.5071428571428571, 'recall': 0.5966386554621849, 'f1': 0.5482625482625483, 'number': 119}	{'precision': 0.9133398247322297, 'recall': 0.8709377901578459, 'f1': 0.8916349809885932, 'number': 1077}	0.8438	0.8644	0.8540	0.7901
0.0141	31.58	600	1.4747	{'precision': 0.8429378531073446, 'recall': 0.9130966952264382, 'f1': 0.8766157461809635, 'number': 817}	{'precision': 0.6493506493506493, 'recall': 0.42016806722689076, 'f1': 0.5102040816326531, 'number': 119}	{'precision': 0.8861788617886179, 'recall': 0.9108635097493036, 'f1': 0.8983516483516484, 'number': 1077}	0.8589	0.8828	0.8707	0.7950
0.0079	42.11	800	1.6605	{'precision': 0.8132464712269273, 'recall': 0.9167686658506732, 'f1': 0.8619102416570771, 'number': 817}	{'precision': 0.5407407407407407, 'recall': 0.6134453781512605, 'f1': 0.5748031496062992, 'number': 119}	{'precision': 0.8919925512104283, 'recall': 0.8895078922934077, 'f1': 0.8907484890748489, 'number': 1077}	0.8357	0.8843	0.8593	0.7910
0.0064	52.63	1000	1.6328	{'precision': 0.8551564310544612, 'recall': 0.9033047735618115, 'f1': 0.8785714285714284, 'number': 817}	{'precision': 0.7466666666666667, 'recall': 0.47058823529411764, 'f1': 0.5773195876288659, 'number': 119}	{'precision': 0.9029918404351768, 'recall': 0.924791086350975, 'f1': 0.9137614678899083, 'number': 1077}	0.8770	0.8892	0.8831	0.7950
0.0091	63.16	1200	1.6620	{'precision': 0.8280044101433297, 'recall': 0.9192166462668299, 'f1': 0.87122969837587, 'number': 817}	{'precision': 0.6875, 'recall': 0.46218487394957986, 'f1': 0.5527638190954773, 'number': 119}	{'precision': 0.9016697588126159, 'recall': 0.9025069637883009, 'f1': 0.9020881670533641, 'number': 1077}	0.8610	0.8833	0.8720	0.7979
0.0031	73.68	1400	1.7310	{'precision': 0.855039637599094, 'recall': 0.9241126070991432, 'f1': 0.8882352941176471, 'number': 817}	{'precision': 0.7023809523809523, 'recall': 0.4957983193277311, 'f1': 0.58128078817734, 'number': 119}	{'precision': 0.8969917958067457, 'recall': 0.9136490250696379, 'f1': 0.9052437902483901, 'number': 1077}	0.8711	0.8932	0.8820	0.7938
0.002	84.21	1600	1.6978	{'precision': 0.870023419203747, 'recall': 0.9094247246022031, 'f1': 0.8892878515858766, 'number': 817}	{'precision': 0.6987951807228916, 'recall': 0.48739495798319327, 'f1': 0.5742574257425742, 'number': 119}	{'precision': 0.9021237303785781, 'recall': 0.9071494893221913, 'f1': 0.9046296296296297, 'number': 1077}	0.8802	0.8833	0.8817	0.8048
0.0023	94.74	1800	1.5996	{'precision': 0.8639534883720931, 'recall': 0.9094247246022031, 'f1': 0.8861061419200955, 'number': 817}	{'precision': 0.6629213483146067, 'recall': 0.4957983193277311, 'f1': 0.5673076923076922, 'number': 119}	{'precision': 0.8969641214351426, 'recall': 0.9052924791086351, 'f1': 0.9011090573012939, 'number': 1077}	0.8728	0.8828	0.8777	0.8026
0.0009	105.26	2000	1.7239	{'precision': 0.8851674641148325, 'recall': 0.9057527539779682, 'f1': 0.8953418027828192, 'number': 817}	{'precision': 0.5726495726495726, 'recall': 0.5630252100840336, 'f1': 0.5677966101694915, 'number': 119}	{'precision': 0.889487870619946, 'recall': 0.9192200557103064, 'f1': 0.9041095890410958, 'number': 1077}	0.8698	0.8927	0.8811	0.8008
0.0009	115.79	2200	1.6091	{'precision': 0.8576349024110218, 'recall': 0.9143206854345165, 'f1': 0.8850710900473934, 'number': 817}	{'precision': 0.6213592233009708, 'recall': 0.5378151260504201, 'f1': 0.5765765765765765, 'number': 119}	{'precision': 0.8901098901098901, 'recall': 0.9025069637883009, 'f1': 0.896265560165975, 'number': 1077}	0.8630	0.8857	0.8742	0.8093
0.0003	126.32	2400	1.7210	{'precision': 0.8627450980392157, 'recall': 0.9155446756425949, 'f1': 0.8883610451306414, 'number': 817}	{'precision': 0.6555555555555556, 'recall': 0.4957983193277311, 'f1': 0.5645933014354068, 'number': 119}	{'precision': 0.8873994638069705, 'recall': 0.9220055710306406, 'f1': 0.9043715846994536, 'number': 1077}	0.8671	0.8942	0.8804	0.8072
0.0006	136.84	2600	1.8215	{'precision': 0.8293478260869566, 'recall': 0.9339045287637698, 'f1': 0.8785261945883708, 'number': 817}	{'precision': 0.5887850467289719, 'recall': 0.5294117647058824, 'f1': 0.5575221238938053, 'number': 119}	{'precision': 0.9232227488151659, 'recall': 0.904363974001857, 'f1': 0.9136960600375235, 'number': 1077}	0.8646	0.8942	0.8791	0.7946
0.0002	147.37	2800	1.8084	{'precision': 0.875886524822695, 'recall': 0.9069767441860465, 'f1': 0.8911605532170775, 'number': 817}	{'precision': 0.5752212389380531, 'recall': 0.5462184873949579, 'f1': 0.5603448275862069, 'number': 119}	{'precision': 0.8896860986547085, 'recall': 0.9210770659238626, 'f1': 0.9051094890510949, 'number': 1077}	0.8669	0.8932	0.8799	0.8000
0.0002	157.89	3000	1.8224	{'precision': 0.9067164179104478, 'recall': 0.8922888616891065, 'f1': 0.8994447871684145, 'number': 817}	{'precision': 0.6222222222222222, 'recall': 0.47058823529411764, 'f1': 0.5358851674641149, 'number': 119}	{'precision': 0.8937112488928255, 'recall': 0.9368616527390901, 'f1': 0.9147778785131461, 'number': 1077}	0.8868	0.8912	0.8890	0.8082
0.0005	168.42	3200	1.7383	{'precision': 0.8754406580493537, 'recall': 0.9118727050183598, 'f1': 0.8932853717026379, 'number': 817}	{'precision': 0.6451612903225806, 'recall': 0.5042016806722689, 'f1': 0.5660377358490566, 'number': 119}	{'precision': 0.8995475113122172, 'recall': 0.9229340761374187, 'f1': 0.9110907424381303, 'number': 1077}	0.8780	0.8937	0.8858	0.8117
0.0002	178.95	3400	1.7757	{'precision': 0.885954381752701, 'recall': 0.9033047735618115, 'f1': 0.8945454545454545, 'number': 817}	{'precision': 0.6896551724137931, 'recall': 0.5042016806722689, 'f1': 0.5825242718446602, 'number': 119}	{'precision': 0.9043321299638989, 'recall': 0.9303621169916435, 'f1': 0.917162471395881, 'number': 1077}	0.8876	0.8942	0.8909	0.8104
0.0001	189.47	3600	1.7467	{'precision': 0.8645833333333334, 'recall': 0.9143206854345165, 'f1': 0.8887566924449734, 'number': 817}	{'precision': 0.6354166666666666, 'recall': 0.5126050420168067, 'f1': 0.5674418604651162, 'number': 119}	{'precision': 0.9027522935779817, 'recall': 0.9136490250696379, 'f1': 0.9081679741578219, 'number': 1077}	0.8741	0.8902	0.8821	0.8054
0.0002	200.0	3800	1.7730	{'precision': 0.8631090487238979, 'recall': 0.9106487148102815, 'f1': 0.8862418106015486, 'number': 817}	{'precision': 0.6354166666666666, 'recall': 0.5126050420168067, 'f1': 0.5674418604651162, 'number': 119}	{'precision': 0.8952205882352942, 'recall': 0.904363974001857, 'f1': 0.899769053117783, 'number': 1077}	0.8695	0.8838	0.8766	0.8024
0.0001	210.53	4000	1.7735	{'precision': 0.8693115519253208, 'recall': 0.9118727050183598, 'f1': 0.8900836320191159, 'number': 817}	{'precision': 0.6630434782608695, 'recall': 0.5126050420168067, 'f1': 0.5781990521327014, 'number': 119}	{'precision': 0.8992673992673993, 'recall': 0.9117920148560817, 'f1': 0.905486399262333, 'number': 1077}	0.8760	0.8882	0.8821	0.8027

Framework versions

• Transformers 4.24.0
• Pytorch 1.12.1
• Datasets 2.7.1
• Tokenizers 0.13.2