layoutlm-finetuned-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: 0.6806
• Answer: {'precision': 0.7069154774972558, 'recall': 0.796044499381953, 'f1': 0.7488372093023256, 'number': 809}
• Header: {'precision': 0.36752136752136755, 'recall': 0.36134453781512604, 'f1': 0.3644067796610169, 'number': 119}
• Question: {'precision': 0.7866549604916594, 'recall': 0.8413145539906103, 'f1': 0.8130671506352087, 'number': 1065}
• Overall Precision: 0.7305
• Overall Recall: 0.7943
• Overall F1: 0.7611
• Overall Accuracy: 0.8085

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
1.8023	1.0	10	1.6152	{'precision': 0.009448818897637795, 'recall': 0.007416563658838072, 'f1': 0.008310249307479225, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.12998266897746968, 'recall': 0.07042253521126761, 'f1': 0.09135200974421437, 'number': 1065}	0.0668	0.0406	0.0505	0.3276
1.4614	2.0	20	1.2547	{'precision': 0.2211253701875617, 'recall': 0.276885043263288, 'f1': 0.24588364434687154, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.43353028064992616, 'recall': 0.5511737089201878, 'f1': 0.4853245142620918, 'number': 1065}	0.3420	0.4069	0.3717	0.5926
1.0795	3.0	30	0.9150	{'precision': 0.49837486457204766, 'recall': 0.5686032138442522, 'f1': 0.5311778290993071, 'number': 809}	{'precision': 0.04081632653061224, 'recall': 0.01680672268907563, 'f1': 0.023809523809523808, 'number': 119}	{'precision': 0.5953338696701529, 'recall': 0.6948356807511737, 'f1': 0.6412478336221837, 'number': 1065}	0.5427	0.6031	0.5713	0.7145
0.8025	4.0	40	0.7686	{'precision': 0.6056622851365016, 'recall': 0.7404202719406675, 'f1': 0.6662958843159066, 'number': 809}	{'precision': 0.10975609756097561, 'recall': 0.07563025210084033, 'f1': 0.08955223880597014, 'number': 119}	{'precision': 0.6737400530503979, 'recall': 0.7154929577464789, 'f1': 0.6939890710382515, 'number': 1065}	0.6222	0.6874	0.6532	0.7506
0.6638	5.0	50	0.7034	{'precision': 0.644535240040858, 'recall': 0.7799752781211372, 'f1': 0.7058165548098434, 'number': 809}	{'precision': 0.23711340206185566, 'recall': 0.19327731092436976, 'f1': 0.21296296296296294, 'number': 119}	{'precision': 0.7158992180712423, 'recall': 0.7737089201877935, 'f1': 0.7436823104693141, 'number': 1065}	0.6637	0.7416	0.7005	0.7841
0.5567	6.0	60	0.6784	{'precision': 0.6687435098650052, 'recall': 0.796044499381953, 'f1': 0.7268623024830699, 'number': 809}	{'precision': 0.2857142857142857, 'recall': 0.2184873949579832, 'f1': 0.24761904761904763, 'number': 119}	{'precision': 0.7160392798690671, 'recall': 0.8215962441314554, 'f1': 0.765194578049847, 'number': 1065}	0.6788	0.7752	0.7238	0.7903
0.4925	7.0	70	0.6815	{'precision': 0.6839779005524862, 'recall': 0.765142150803461, 'f1': 0.7222870478413069, 'number': 809}	{'precision': 0.2894736842105263, 'recall': 0.2773109243697479, 'f1': 0.2832618025751073, 'number': 119}	{'precision': 0.7233169129720853, 'recall': 0.8272300469483568, 'f1': 0.7717915024091108, 'number': 1065}	0.6853	0.7692	0.7248	0.7913
0.4494	8.0	80	0.6765	{'precision': 0.6962305986696231, 'recall': 0.7762669962917181, 'f1': 0.734073641145529, 'number': 809}	{'precision': 0.28, 'recall': 0.29411764705882354, 'f1': 0.28688524590163933, 'number': 119}	{'precision': 0.7360066833751044, 'recall': 0.8272300469483568, 'f1': 0.7789566755083996, 'number': 1065}	0.6942	0.7747	0.7323	0.8004
0.3986	9.0	90	0.6587	{'precision': 0.7077777777777777, 'recall': 0.7873918417799752, 'f1': 0.7454651843183148, 'number': 809}	{'precision': 0.3274336283185841, 'recall': 0.31092436974789917, 'f1': 0.3189655172413793, 'number': 119}	{'precision': 0.7487266553480475, 'recall': 0.828169014084507, 'f1': 0.7864467231386535, 'number': 1065}	0.7102	0.7807	0.7438	0.8019
0.3597	10.0	100	0.6607	{'precision': 0.7054945054945055, 'recall': 0.7935723114956736, 'f1': 0.7469458987783596, 'number': 809}	{'precision': 0.3305084745762712, 'recall': 0.3277310924369748, 'f1': 0.32911392405063294, 'number': 119}	{'precision': 0.7600685518423308, 'recall': 0.8328638497652582, 'f1': 0.7948028673835125, 'number': 1065}	0.7144	0.7868	0.7488	0.8048
0.3266	11.0	110	0.6751	{'precision': 0.7050279329608938, 'recall': 0.7799752781211372, 'f1': 0.7406103286384977, 'number': 809}	{'precision': 0.3684210526315789, 'recall': 0.35294117647058826, 'f1': 0.3605150214592275, 'number': 119}	{'precision': 0.7711571675302246, 'recall': 0.8384976525821596, 'f1': 0.8034188034188035, 'number': 1065}	0.7227	0.7858	0.7529	0.8056
0.3103	12.0	120	0.6799	{'precision': 0.7047413793103449, 'recall': 0.8084054388133498, 'f1': 0.7530224525043179, 'number': 809}	{'precision': 0.3474576271186441, 'recall': 0.3445378151260504, 'f1': 0.3459915611814346, 'number': 119}	{'precision': 0.7799295774647887, 'recall': 0.831924882629108, 'f1': 0.8050885960926851, 'number': 1065}	0.7246	0.7933	0.7574	0.8034
0.2893	13.0	130	0.6773	{'precision': 0.7191392978482446, 'recall': 0.7849196538936959, 'f1': 0.7505910165484633, 'number': 809}	{'precision': 0.35833333333333334, 'recall': 0.36134453781512604, 'f1': 0.35983263598326365, 'number': 119}	{'precision': 0.7861524978089395, 'recall': 0.8422535211267606, 'f1': 0.8132366273798729, 'number': 1065}	0.7346	0.7903	0.7614	0.8050
0.2743	14.0	140	0.6788	{'precision': 0.7063318777292577, 'recall': 0.799752781211372, 'f1': 0.750144927536232, 'number': 809}	{'precision': 0.37168141592920356, 'recall': 0.35294117647058826, 'f1': 0.36206896551724144, 'number': 119}	{'precision': 0.7879858657243817, 'recall': 0.8375586854460094, 'f1': 0.812016385980883, 'number': 1065}	0.7316	0.7933	0.7612	0.8085
0.2756	15.0	150	0.6806	{'precision': 0.7069154774972558, 'recall': 0.796044499381953, 'f1': 0.7488372093023256, 'number': 809}	{'precision': 0.36752136752136755, 'recall': 0.36134453781512604, 'f1': 0.3644067796610169, 'number': 119}	{'precision': 0.7866549604916594, 'recall': 0.8413145539906103, 'f1': 0.8130671506352087, 'number': 1065}	0.7305	0.7943	0.7611	0.8085

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.1+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2