layoutlm-funsd

This model is a fine-tuned version of microsoft/layoutlm-base-uncased on the funsd dataset. It achieves the following results on the evaluation set:

• Loss: 0.6724
• Answer: {'precision': 0.7072368421052632, 'recall': 0.7972805933250927, 'f1': 0.7495642068564788, 'number': 809}
• Header: {'precision': 0.3333333333333333, 'recall': 0.3697478991596639, 'f1': 0.350597609561753, 'number': 119}
• Question: {'precision': 0.7901785714285714, 'recall': 0.8309859154929577, 'f1': 0.8100686498855836, 'number': 1065}
• Overall Precision: 0.7274
• Overall Recall: 0.7898
• Overall F1: 0.7573
• Overall Accuracy: 0.8170

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: 3e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Answer	Header	Question	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
1.7705	1.0	10	1.5739	{'precision': 0.010057471264367816, 'recall': 0.00865265760197775, 'f1': 0.00930232558139535, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.26591760299625467, 'recall': 0.13333333333333333, 'f1': 0.17761100687929957, 'number': 1065}	0.1211	0.0748	0.0925	0.3598
1.4468	2.0	20	1.2327	{'precision': 0.25151148730350664, 'recall': 0.25710754017305315, 'f1': 0.2542787286063569, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.4251316779533484, 'recall': 0.5305164319248826, 'f1': 0.4720133667502089, 'number': 1065}	0.3585	0.3879	0.3726	0.5921
1.1103	3.0	30	0.9608	{'precision': 0.4880694143167028, 'recall': 0.5562422744128553, 'f1': 0.5199306759098786, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.5613598673300166, 'recall': 0.6356807511737089, 'f1': 0.5962131219726994, 'number': 1065}	0.5257	0.5655	0.5448	0.7058
0.8476	4.0	40	0.7875	{'precision': 0.5821325648414986, 'recall': 0.7490729295426453, 'f1': 0.6551351351351351, 'number': 809}	{'precision': 0.1702127659574468, 'recall': 0.06722689075630252, 'f1': 0.09638554216867469, 'number': 119}	{'precision': 0.6254071661237784, 'recall': 0.7211267605633803, 'f1': 0.6698648059310947, 'number': 1065}	0.5967	0.6934	0.6414	0.7538
0.6698	5.0	50	0.6948	{'precision': 0.6421923474663909, 'recall': 0.7676143386897404, 'f1': 0.6993243243243243, 'number': 809}	{'precision': 0.3424657534246575, 'recall': 0.21008403361344538, 'f1': 0.2604166666666667, 'number': 119}	{'precision': 0.6873977086743044, 'recall': 0.7887323943661971, 'f1': 0.7345867949278532, 'number': 1065}	0.6569	0.7456	0.6985	0.7860
0.554	6.0	60	0.6717	{'precision': 0.6506410256410257, 'recall': 0.7527812113720643, 'f1': 0.6979942693409743, 'number': 809}	{'precision': 0.3448275862068966, 'recall': 0.25210084033613445, 'f1': 0.2912621359223301, 'number': 119}	{'precision': 0.716821639898563, 'recall': 0.7962441314553991, 'f1': 0.7544483985765124, 'number': 1065}	0.6741	0.7461	0.7083	0.7920
0.4787	7.0	70	0.6462	{'precision': 0.6666666666666666, 'recall': 0.7935723114956736, 'f1': 0.7246049661399547, 'number': 809}	{'precision': 0.3017241379310345, 'recall': 0.29411764705882354, 'f1': 0.29787234042553185, 'number': 119}	{'precision': 0.7368421052631579, 'recall': 0.8018779342723005, 'f1': 0.7679856115107914, 'number': 1065}	0.6841	0.7682	0.7237	0.8038
0.4182	8.0	80	0.6516	{'precision': 0.6790890269151139, 'recall': 0.8108776266996292, 'f1': 0.7391549295774646, 'number': 809}	{'precision': 0.3063063063063063, 'recall': 0.2857142857142857, 'f1': 0.2956521739130435, 'number': 119}	{'precision': 0.7450643776824034, 'recall': 0.8150234741784037, 'f1': 0.77847533632287, 'number': 1065}	0.6949	0.7817	0.7358	0.8025
0.3877	9.0	90	0.6652	{'precision': 0.6976744186046512, 'recall': 0.7787391841779975, 'f1': 0.7359813084112149, 'number': 809}	{'precision': 0.3194444444444444, 'recall': 0.3865546218487395, 'f1': 0.34980988593155893, 'number': 119}	{'precision': 0.7573913043478261, 'recall': 0.8178403755868544, 'f1': 0.7864559819413092, 'number': 1065}	0.7041	0.7762	0.7384	0.8094
0.3483	10.0	100	0.6568	{'precision': 0.6876332622601279, 'recall': 0.7972805933250927, 'f1': 0.7384087006296507, 'number': 809}	{'precision': 0.3225806451612903, 'recall': 0.33613445378151263, 'f1': 0.3292181069958848, 'number': 119}	{'precision': 0.7650602409638554, 'recall': 0.8347417840375587, 'f1': 0.7983834755276155, 'number': 1065}	0.7077	0.7898	0.7465	0.8151
0.3136	11.0	110	0.6698	{'precision': 0.7006507592190889, 'recall': 0.7985166872682324, 'f1': 0.7463893703061815, 'number': 809}	{'precision': 0.3247863247863248, 'recall': 0.31932773109243695, 'f1': 0.3220338983050848, 'number': 119}	{'precision': 0.7803365810451727, 'recall': 0.8272300469483568, 'f1': 0.8030993618960802, 'number': 1065}	0.7219	0.7852	0.7522	0.8084
0.3044	12.0	120	0.6667	{'precision': 0.7058177826564215, 'recall': 0.7948084054388134, 'f1': 0.7476744186046511, 'number': 809}	{'precision': 0.34328358208955223, 'recall': 0.3865546218487395, 'f1': 0.36363636363636365, 'number': 119}	{'precision': 0.785204991087344, 'recall': 0.8272300469483568, 'f1': 0.8056698673982624, 'number': 1065}	0.7245	0.7878	0.7548	0.8138
0.2853	13.0	130	0.6699	{'precision': 0.702819956616052, 'recall': 0.8009888751545118, 'f1': 0.7487001733102253, 'number': 809}	{'precision': 0.3442622950819672, 'recall': 0.35294117647058826, 'f1': 0.3485477178423237, 'number': 119}	{'precision': 0.7857769973661106, 'recall': 0.8403755868544601, 'f1': 0.8121597096188747, 'number': 1065}	0.7261	0.7953	0.7591	0.8136
0.278	14.0	140	0.6716	{'precision': 0.7009750812567714, 'recall': 0.799752781211372, 'f1': 0.7471131639722863, 'number': 809}	{'precision': 0.3233082706766917, 'recall': 0.36134453781512604, 'f1': 0.3412698412698413, 'number': 119}	{'precision': 0.7873665480427047, 'recall': 0.8309859154929577, 'f1': 0.808588396528095, 'number': 1065}	0.7225	0.7903	0.7549	0.8154
0.2672	15.0	150	0.6724	{'precision': 0.7072368421052632, 'recall': 0.7972805933250927, 'f1': 0.7495642068564788, 'number': 809}	{'precision': 0.3333333333333333, 'recall': 0.3697478991596639, 'f1': 0.350597609561753, 'number': 119}	{'precision': 0.7901785714285714, 'recall': 0.8309859154929577, 'f1': 0.8100686498855836, 'number': 1065}	0.7274	0.7898	0.7573	0.8170

Framework versions

• Transformers 4.25.1
• Pytorch 1.9.0
• Datasets 2.8.0
• Tokenizers 0.13.2