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: 0.6771
• Answer: {'precision': 0.7053571428571429, 'recall': 0.7812113720642769, 'f1': 0.7413489736070381, 'number': 809}
• Header: {'precision': 0.29770992366412213, 'recall': 0.3277310924369748, 'f1': 0.312, 'number': 119}
• Question: {'precision': 0.763716814159292, 'recall': 0.8103286384976526, 'f1': 0.7863325740318907, 'number': 1065}
• Overall Precision: 0.7112
• Overall Recall: 0.7697
• Overall F1: 0.7393
• Overall Accuracy: 0.8036

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
1.7668	1.0	10	1.5655	{'precision': 0.01564945226917058, 'recall': 0.012360939431396786, 'f1': 0.013812154696132598, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.31925675675675674, 'recall': 0.17746478873239438, 'f1': 0.22812311406155703, 'number': 1065}	0.1617	0.0998	0.1234	0.3599
1.4335	2.0	20	1.2504	{'precision': 0.22972972972972974, 'recall': 0.2521631644004944, 'f1': 0.24042427813789036, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.391304347826087, 'recall': 0.5492957746478874, 'f1': 0.45703125, 'number': 1065}	0.3311	0.3959	0.3606	0.5926
1.1215	3.0	30	0.9771	{'precision': 0.4570273003033367, 'recall': 0.5587144622991347, 'f1': 0.5027808676307008, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.5460829493087558, 'recall': 0.6676056338028169, 'f1': 0.6007604562737643, 'number': 1065}	0.5022	0.5835	0.5398	0.6802
0.8632	4.0	40	0.7960	{'precision': 0.5714285714285714, 'recall': 0.7317676143386898, 'f1': 0.6417344173441735, 'number': 809}	{'precision': 0.0379746835443038, 'recall': 0.025210084033613446, 'f1': 0.030303030303030304, 'number': 119}	{'precision': 0.633201581027668, 'recall': 0.752112676056338, 'f1': 0.6875536480686696, 'number': 1065}	0.5866	0.7005	0.6385	0.7572
0.6913	5.0	50	0.7202	{'precision': 0.6335797254487856, 'recall': 0.7416563658838071, 'f1': 0.683371298405467, 'number': 809}	{'precision': 0.10752688172043011, 'recall': 0.08403361344537816, 'f1': 0.09433962264150944, 'number': 119}	{'precision': 0.6907216494845361, 'recall': 0.7549295774647887, 'f1': 0.721399730820996, 'number': 1065}	0.6416	0.7095	0.6738	0.7808
0.5758	6.0	60	0.6812	{'precision': 0.6677248677248677, 'recall': 0.7799752781211372, 'f1': 0.7194982896237172, 'number': 809}	{'precision': 0.17857142857142858, 'recall': 0.16806722689075632, 'f1': 0.17316017316017318, 'number': 119}	{'precision': 0.6919315403422983, 'recall': 0.7971830985915493, 'f1': 0.7408376963350787, 'number': 1065}	0.6567	0.7526	0.7014	0.7947
0.4993	7.0	70	0.6593	{'precision': 0.6597938144329897, 'recall': 0.7911001236093943, 'f1': 0.7195053400786959, 'number': 809}	{'precision': 0.21875, 'recall': 0.17647058823529413, 'f1': 0.19534883720930232, 'number': 119}	{'precision': 0.7158703071672355, 'recall': 0.787793427230047, 'f1': 0.7501117568171659, 'number': 1065}	0.6702	0.7526	0.7091	0.7951
0.4397	8.0	80	0.6633	{'precision': 0.6749192680301399, 'recall': 0.7750309023485785, 'f1': 0.721518987341772, 'number': 809}	{'precision': 0.23943661971830985, 'recall': 0.2857142857142857, 'f1': 0.26053639846743293, 'number': 119}	{'precision': 0.7274290627687017, 'recall': 0.7943661971830986, 'f1': 0.7594254937163376, 'number': 1065}	0.6746	0.7561	0.7130	0.7987
0.396	9.0	90	0.6605	{'precision': 0.6875699888017918, 'recall': 0.7589616810877626, 'f1': 0.7215041128084607, 'number': 809}	{'precision': 0.25203252032520324, 'recall': 0.2605042016806723, 'f1': 0.25619834710743805, 'number': 119}	{'precision': 0.7344013490725126, 'recall': 0.8178403755868544, 'f1': 0.7738782763216348, 'number': 1065}	0.6885	0.7607	0.7228	0.8026
0.3578	10.0	100	0.6624	{'precision': 0.6842105263157895, 'recall': 0.7873918417799752, 'f1': 0.7321839080459769, 'number': 809}	{'precision': 0.2773109243697479, 'recall': 0.2773109243697479, 'f1': 0.2773109243697479, 'number': 119}	{'precision': 0.7476231633535004, 'recall': 0.812206572769953, 'f1': 0.7785778577857785, 'number': 1065}	0.6955	0.7702	0.7310	0.8022
0.3249	11.0	110	0.6678	{'precision': 0.6865671641791045, 'recall': 0.796044499381953, 'f1': 0.7372638809387521, 'number': 809}	{'precision': 0.273972602739726, 'recall': 0.33613445378151263, 'f1': 0.3018867924528302, 'number': 119}	{'precision': 0.7584973166368515, 'recall': 0.7962441314553991, 'f1': 0.7769125057260651, 'number': 1065}	0.6957	0.7687	0.7304	0.8010
0.3021	12.0	120	0.6681	{'precision': 0.7120535714285714, 'recall': 0.788627935723115, 'f1': 0.7483870967741936, 'number': 809}	{'precision': 0.2923076923076923, 'recall': 0.31932773109243695, 'f1': 0.3052208835341365, 'number': 119}	{'precision': 0.7547826086956522, 'recall': 0.8150234741784037, 'f1': 0.7837471783295711, 'number': 1065}	0.7096	0.7747	0.7407	0.8045
0.2895	13.0	130	0.6755	{'precision': 0.7122060470324748, 'recall': 0.7861557478368356, 'f1': 0.7473560517038778, 'number': 809}	{'precision': 0.2887323943661972, 'recall': 0.3445378151260504, 'f1': 0.31417624521072796, 'number': 119}	{'precision': 0.7595048629531388, 'recall': 0.8065727699530516, 'f1': 0.7823315118397085, 'number': 1065}	0.7091	0.7707	0.7386	0.8026
0.2734	14.0	140	0.6768	{'precision': 0.7093541202672605, 'recall': 0.7873918417799752, 'f1': 0.7463386057410661, 'number': 809}	{'precision': 0.28368794326241137, 'recall': 0.33613445378151263, 'f1': 0.3076923076923077, 'number': 119}	{'precision': 0.7570175438596491, 'recall': 0.8103286384976526, 'f1': 0.7827664399092971, 'number': 1065}	0.7067	0.7727	0.7383	0.8026
0.2804	15.0	150	0.6771	{'precision': 0.7053571428571429, 'recall': 0.7812113720642769, 'f1': 0.7413489736070381, 'number': 809}	{'precision': 0.29770992366412213, 'recall': 0.3277310924369748, 'f1': 0.312, 'number': 119}	{'precision': 0.763716814159292, 'recall': 0.8103286384976526, 'f1': 0.7863325740318907, 'number': 1065}	0.7112	0.7697	0.7393	0.8036

Framework versions

• Transformers 4.34.0
• Pytorch 2.1.0+cpu
• Datasets 2.14.5
• Tokenizers 0.14.1