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: 0.6820
• Answer: {'precision': 0.7084257206208425, 'recall': 0.7898640296662547, 'f1': 0.7469316189362946, 'number': 809}
• Header: {'precision': 0.2689655172413793, 'recall': 0.3277310924369748, 'f1': 0.2954545454545454, 'number': 119}
• Question: {'precision': 0.7870619946091644, 'recall': 0.8225352112676056, 'f1': 0.8044077134986226, 'number': 1065}
• Overall Precision: 0.7194
• Overall Recall: 0.7797
• Overall F1: 0.7484
• Overall Accuracy: 0.8102

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
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Answer	Header	Question	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
1.7857	1.0	10	1.5985	{'precision': 0.009248554913294798, 'recall': 0.009888751545117428, 'f1': 0.00955794504181601, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.1273972602739726, 'recall': 0.08732394366197183, 'f1': 0.10362116991643454, 'number': 1065}	0.0633	0.0507	0.0563	0.3562
1.4597	2.0	20	1.2331	{'precision': 0.18717683557394002, 'recall': 0.22373300370828184, 'f1': 0.20382882882882883, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.4439461883408072, 'recall': 0.5577464788732395, 'f1': 0.4943820224719101, 'number': 1065}	0.3362	0.3889	0.3606	0.6007
1.0902	3.0	30	0.9489	{'precision': 0.4371069182389937, 'recall': 0.515451174289246, 'f1': 0.47305728871242203, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.6257615317667538, 'recall': 0.6751173708920187, 'f1': 0.6495031616982836, 'number': 1065}	0.5311	0.5700	0.5499	0.6910
0.8339	4.0	40	0.7979	{'precision': 0.5977366255144033, 'recall': 0.7181705809641533, 'f1': 0.652442448062886, 'number': 809}	{'precision': 0.13513513513513514, 'recall': 0.08403361344537816, 'f1': 0.10362694300518135, 'number': 119}	{'precision': 0.6854545454545454, 'recall': 0.707981220657277, 'f1': 0.6965357967667436, 'number': 1065}	0.6267	0.6749	0.6499	0.7453
0.6983	5.0	50	0.7659	{'precision': 0.6392896781354052, 'recall': 0.7119901112484549, 'f1': 0.6736842105263159, 'number': 809}	{'precision': 0.19626168224299065, 'recall': 0.17647058823529413, 'f1': 0.18584070796460178, 'number': 119}	{'precision': 0.6688102893890675, 'recall': 0.7812206572769953, 'f1': 0.7206582936336077, 'number': 1065}	0.6345	0.7170	0.6733	0.7610
0.5815	6.0	60	0.6907	{'precision': 0.6410256410256411, 'recall': 0.7725587144622992, 'f1': 0.7006726457399104, 'number': 809}	{'precision': 0.23863636363636365, 'recall': 0.17647058823529413, 'f1': 0.20289855072463767, 'number': 119}	{'precision': 0.7027463651050081, 'recall': 0.8169014084507042, 'f1': 0.7555362570560139, 'number': 1065}	0.6588	0.7607	0.7061	0.7913
0.5044	7.0	70	0.6802	{'precision': 0.6727078891257996, 'recall': 0.7799752781211372, 'f1': 0.7223812249570692, 'number': 809}	{'precision': 0.26605504587155965, 'recall': 0.24369747899159663, 'f1': 0.2543859649122807, 'number': 119}	{'precision': 0.7305699481865285, 'recall': 0.7943661971830986, 'f1': 0.7611336032388665, 'number': 1065}	0.6830	0.7556	0.7175	0.7902
0.4534	8.0	80	0.6595	{'precision': 0.7018701870187019, 'recall': 0.788627935723115, 'f1': 0.7427240977881256, 'number': 809}	{'precision': 0.234375, 'recall': 0.25210084033613445, 'f1': 0.242914979757085, 'number': 119}	{'precision': 0.7378559463986599, 'recall': 0.8272300469483568, 'f1': 0.779991146525011, 'number': 1065}	0.6943	0.7772	0.7334	0.8074
0.3971	9.0	90	0.6625	{'precision': 0.6967032967032967, 'recall': 0.7836835599505563, 'f1': 0.7376381617219313, 'number': 809}	{'precision': 0.27007299270072993, 'recall': 0.31092436974789917, 'f1': 0.2890625, 'number': 119}	{'precision': 0.7433930093776641, 'recall': 0.8187793427230047, 'f1': 0.7792672028596961, 'number': 1065}	0.6950	0.7742	0.7325	0.8060
0.3593	10.0	100	0.6634	{'precision': 0.7079152731326644, 'recall': 0.7849196538936959, 'f1': 0.7444314185228605, 'number': 809}	{'precision': 0.2714285714285714, 'recall': 0.31932773109243695, 'f1': 0.29343629343629346, 'number': 119}	{'precision': 0.7571305099394987, 'recall': 0.8225352112676056, 'f1': 0.7884788478847885, 'number': 1065}	0.7060	0.7772	0.7399	0.8115
0.3209	11.0	110	0.6655	{'precision': 0.6973262032085561, 'recall': 0.8059332509270705, 'f1': 0.7477064220183487, 'number': 809}	{'precision': 0.2903225806451613, 'recall': 0.3025210084033613, 'f1': 0.2962962962962963, 'number': 119}	{'precision': 0.7788632326820604, 'recall': 0.8234741784037559, 'f1': 0.8005476951163851, 'number': 1065}	0.7162	0.7852	0.7492	0.8129
0.3064	12.0	120	0.6771	{'precision': 0.7104072398190046, 'recall': 0.7762669962917181, 'f1': 0.74187832250443, 'number': 809}	{'precision': 0.2689655172413793, 'recall': 0.3277310924369748, 'f1': 0.2954545454545454, 'number': 119}	{'precision': 0.7795698924731183, 'recall': 0.8169014084507042, 'f1': 0.797799174690509, 'number': 1065}	0.7166	0.7712	0.7429	0.8088
0.286	13.0	130	0.6765	{'precision': 0.7030905077262694, 'recall': 0.7873918417799752, 'f1': 0.7428571428571429, 'number': 809}	{'precision': 0.2689655172413793, 'recall': 0.3277310924369748, 'f1': 0.2954545454545454, 'number': 119}	{'precision': 0.769298245614035, 'recall': 0.8234741784037559, 'f1': 0.7954648526077097, 'number': 1065}	0.7088	0.7792	0.7424	0.8111
0.2806	14.0	140	0.6820	{'precision': 0.7052980132450332, 'recall': 0.7898640296662547, 'f1': 0.7451895043731779, 'number': 809}	{'precision': 0.2689655172413793, 'recall': 0.3277310924369748, 'f1': 0.2954545454545454, 'number': 119}	{'precision': 0.7793594306049823, 'recall': 0.8225352112676056, 'f1': 0.8003654636820466, 'number': 1065}	0.7145	0.7797	0.7457	0.8106
0.2736	15.0	150	0.6820	{'precision': 0.7084257206208425, 'recall': 0.7898640296662547, 'f1': 0.7469316189362946, 'number': 809}	{'precision': 0.2689655172413793, 'recall': 0.3277310924369748, 'f1': 0.2954545454545454, 'number': 119}	{'precision': 0.7870619946091644, 'recall': 0.8225352112676056, 'f1': 0.8044077134986226, 'number': 1065}	0.7194	0.7797	0.7484	0.8102

Framework versions

• Transformers 4.23.1
• Pytorch 1.12.1+cu113
• Datasets 2.6.1
• Tokenizers 0.13.1