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: 0.8048
• Answer: {'precision': 0.7424412094064949, 'recall': 0.8195302843016069, 'f1': 0.7790834312573444, 'number': 809}
• Header: {'precision': 0.41304347826086957, 'recall': 0.4789915966386555, 'f1': 0.443579766536965, 'number': 119}
• Question: {'precision': 0.8048561151079137, 'recall': 0.8403755868544601, 'f1': 0.8222324299494717, 'number': 1065}
• Overall Precision: 0.7536
• Overall Recall: 0.8103
• Overall F1: 0.7809
• Overall Accuracy: 0.8256

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: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Answer	Header	Question	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
1.8389	1.0	10	1.6291	{'precision': 0.01568627450980392, 'recall': 0.009888751545117428, 'f1': 0.012130401819560273, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.2798165137614679, 'recall': 0.11455399061032864, 'f1': 0.16255829447035308, 'number': 1065}	0.1374	0.0652	0.0885	0.3319
1.4797	2.0	20	1.2835	{'precision': 0.2250740375123396, 'recall': 0.28182941903584674, 'f1': 0.2502744237102085, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.3997214484679666, 'recall': 0.5389671361502347, 'f1': 0.45901639344262296, 'number': 1065}	0.3275	0.4024	0.3611	0.5792
1.1281	3.0	30	0.9324	{'precision': 0.47114375655823715, 'recall': 0.5550061804697157, 'f1': 0.5096481271282634, 'number': 809}	{'precision': 0.06060606060606061, 'recall': 0.01680672268907563, 'f1': 0.02631578947368421, 'number': 119}	{'precision': 0.5470149253731343, 'recall': 0.6882629107981221, 'f1': 0.6095634095634096, 'number': 1065}	0.5090	0.5941	0.5483	0.7008
0.848	4.0	40	0.7620	{'precision': 0.5925563173359452, 'recall': 0.7478368355995055, 'f1': 0.6612021857923498, 'number': 809}	{'precision': 0.17391304347826086, 'recall': 0.10084033613445378, 'f1': 0.12765957446808512, 'number': 119}	{'precision': 0.6578293289146645, 'recall': 0.7455399061032864, 'f1': 0.6989436619718309, 'number': 1065}	0.6143	0.7080	0.6578	0.7624
0.6618	5.0	50	0.6889	{'precision': 0.6424180327868853, 'recall': 0.7750309023485785, 'f1': 0.7025210084033613, 'number': 809}	{'precision': 0.3, 'recall': 0.226890756302521, 'f1': 0.25837320574162675, 'number': 119}	{'precision': 0.6919967663702506, 'recall': 0.8037558685446009, 'f1': 0.7437011294526499, 'number': 1065}	0.6557	0.7577	0.7030	0.7901
0.5475	6.0	60	0.6690	{'precision': 0.654158215010142, 'recall': 0.7972805933250927, 'f1': 0.7186629526462396, 'number': 809}	{'precision': 0.31868131868131866, 'recall': 0.24369747899159663, 'f1': 0.2761904761904762, 'number': 119}	{'precision': 0.7417102966841187, 'recall': 0.7981220657276995, 'f1': 0.7688828584350972, 'number': 1065}	0.6856	0.7647	0.7230	0.7949
0.4641	7.0	70	0.6472	{'precision': 0.6896551724137931, 'recall': 0.7911001236093943, 'f1': 0.7369027058146229, 'number': 809}	{'precision': 0.23529411764705882, 'recall': 0.23529411764705882, 'f1': 0.23529411764705882, 'number': 119}	{'precision': 0.7485131690739167, 'recall': 0.8272300469483568, 'f1': 0.7859054415700267, 'number': 1065}	0.6965	0.7772	0.7346	0.8108
0.3968	8.0	80	0.6603	{'precision': 0.7052518756698821, 'recall': 0.8133498145859085, 'f1': 0.7554535017221584, 'number': 809}	{'precision': 0.26277372262773724, 'recall': 0.3025210084033613, 'f1': 0.28125000000000006, 'number': 119}	{'precision': 0.7734513274336283, 'recall': 0.8206572769953052, 'f1': 0.7963553530751709, 'number': 1065}	0.7127	0.7868	0.7479	0.8117
0.3377	9.0	90	0.6641	{'precision': 0.7273730684326711, 'recall': 0.8145859085290482, 'f1': 0.7685131195335277, 'number': 809}	{'precision': 0.30612244897959184, 'recall': 0.37815126050420167, 'f1': 0.3383458646616541, 'number': 119}	{'precision': 0.7655838454784899, 'recall': 0.8187793427230047, 'f1': 0.7912885662431942, 'number': 1065}	0.7190	0.7908	0.7532	0.8063
0.3159	10.0	100	0.6626	{'precision': 0.7112299465240641, 'recall': 0.8220024721878862, 'f1': 0.7626146788990825, 'number': 809}	{'precision': 0.36666666666666664, 'recall': 0.2773109243697479, 'f1': 0.31578947368421056, 'number': 119}	{'precision': 0.7945945945945946, 'recall': 0.828169014084507, 'f1': 0.8110344827586206, 'number': 1065}	0.7400	0.7928	0.7655	0.8252
0.2565	11.0	110	0.6831	{'precision': 0.706951871657754, 'recall': 0.8170580964153276, 'f1': 0.7580275229357798, 'number': 809}	{'precision': 0.3418803418803419, 'recall': 0.33613445378151263, 'f1': 0.3389830508474576, 'number': 119}	{'precision': 0.7935656836461126, 'recall': 0.8338028169014085, 'f1': 0.8131868131868133, 'number': 1065}	0.7319	0.7973	0.7632	0.8146
0.2326	12.0	120	0.7081	{'precision': 0.7152103559870551, 'recall': 0.8195302843016069, 'f1': 0.7638248847926269, 'number': 809}	{'precision': 0.34375, 'recall': 0.3697478991596639, 'f1': 0.3562753036437247, 'number': 119}	{'precision': 0.7731601731601732, 'recall': 0.8384976525821596, 'f1': 0.8045045045045045, 'number': 1065}	0.7240	0.8028	0.7614	0.8097
0.2064	13.0	130	0.7088	{'precision': 0.7420454545454546, 'recall': 0.8071693448702101, 'f1': 0.773238602723505, 'number': 809}	{'precision': 0.375, 'recall': 0.37815126050420167, 'f1': 0.37656903765690375, 'number': 119}	{'precision': 0.7978628673196795, 'recall': 0.8413145539906103, 'f1': 0.8190127970749542, 'number': 1065}	0.7508	0.7998	0.7745	0.8216
0.1807	14.0	140	0.7149	{'precision': 0.7113289760348583, 'recall': 0.8071693448702101, 'f1': 0.7562246670526924, 'number': 809}	{'precision': 0.373134328358209, 'recall': 0.42016806722689076, 'f1': 0.3952569169960475, 'number': 119}	{'precision': 0.8001800180018002, 'recall': 0.8347417840375587, 'f1': 0.8170955882352942, 'number': 1065}	0.7360	0.7988	0.7661	0.8186
0.1673	15.0	150	0.7429	{'precision': 0.7461988304093568, 'recall': 0.788627935723115, 'f1': 0.766826923076923, 'number': 809}	{'precision': 0.4015151515151515, 'recall': 0.44537815126050423, 'f1': 0.4223107569721115, 'number': 119}	{'precision': 0.8001800180018002, 'recall': 0.8347417840375587, 'f1': 0.8170955882352942, 'number': 1065}	0.7531	0.7928	0.7724	0.8213
0.158	16.0	160	0.7579	{'precision': 0.7352614015572859, 'recall': 0.8170580964153276, 'f1': 0.7740046838407495, 'number': 809}	{'precision': 0.3673469387755102, 'recall': 0.453781512605042, 'f1': 0.406015037593985, 'number': 119}	{'precision': 0.790616854908775, 'recall': 0.8544600938967136, 'f1': 0.8212996389891697, 'number': 1065}	0.7396	0.8154	0.7757	0.8166
0.1407	17.0	170	0.7595	{'precision': 0.7474747474747475, 'recall': 0.823238566131026, 'f1': 0.783529411764706, 'number': 809}	{'precision': 0.424, 'recall': 0.44537815126050423, 'f1': 0.4344262295081967, 'number': 119}	{'precision': 0.8081081081081081, 'recall': 0.8422535211267606, 'f1': 0.8248275862068966, 'number': 1065}	0.7601	0.8108	0.7847	0.8237
0.1277	18.0	180	0.7927	{'precision': 0.7305986696230599, 'recall': 0.8145859085290482, 'f1': 0.7703097603740503, 'number': 809}	{'precision': 0.4140625, 'recall': 0.44537815126050423, 'f1': 0.42914979757085026, 'number': 119}	{'precision': 0.8114233907524931, 'recall': 0.8403755868544601, 'f1': 0.8256457564575646, 'number': 1065}	0.7534	0.8063	0.7790	0.8137
0.1268	19.0	190	0.7819	{'precision': 0.7361894024802705, 'recall': 0.8071693448702101, 'f1': 0.7700471698113207, 'number': 809}	{'precision': 0.4330708661417323, 'recall': 0.46218487394957986, 'f1': 0.4471544715447155, 'number': 119}	{'precision': 0.8028419182948491, 'recall': 0.8488262910798122, 'f1': 0.8251939753537197, 'number': 1065}	0.7533	0.8088	0.7801	0.8216
0.1112	20.0	200	0.7880	{'precision': 0.740782122905028, 'recall': 0.8195302843016069, 'f1': 0.7781690140845071, 'number': 809}	{'precision': 0.4195804195804196, 'recall': 0.5042016806722689, 'f1': 0.4580152671755725, 'number': 119}	{'precision': 0.8075880758807588, 'recall': 0.8394366197183099, 'f1': 0.8232044198895027, 'number': 1065}	0.7538	0.8113	0.7815	0.8229
0.1096	21.0	210	0.7925	{'precision': 0.7404494382022472, 'recall': 0.8145859085290482, 'f1': 0.7757504414361388, 'number': 809}	{'precision': 0.45454545454545453, 'recall': 0.42016806722689076, 'f1': 0.43668122270742354, 'number': 119}	{'precision': 0.815049864007253, 'recall': 0.844131455399061, 'f1': 0.8293357933579335, 'number': 1065}	0.7646	0.8068	0.7852	0.8249
0.1158	22.0	220	0.8093	{'precision': 0.7363128491620111, 'recall': 0.8145859085290482, 'f1': 0.7734741784037558, 'number': 809}	{'precision': 0.41333333333333333, 'recall': 0.5210084033613446, 'f1': 0.4609665427509294, 'number': 119}	{'precision': 0.8030438675022381, 'recall': 0.8422535211267606, 'f1': 0.8221814848762603, 'number': 1065}	0.7484	0.8118	0.7788	0.8210
0.0985	23.0	230	0.8013	{'precision': 0.7554535017221584, 'recall': 0.8133498145859085, 'f1': 0.7833333333333333, 'number': 809}	{'precision': 0.45689655172413796, 'recall': 0.44537815126050423, 'f1': 0.4510638297872341, 'number': 119}	{'precision': 0.8091809180918091, 'recall': 0.844131455399061, 'f1': 0.8262867647058824, 'number': 1065}	0.7674	0.8078	0.7871	0.8279
0.0988	24.0	240	0.8040	{'precision': 0.7385984427141268, 'recall': 0.8207663782447466, 'f1': 0.7775175644028104, 'number': 809}	{'precision': 0.4198473282442748, 'recall': 0.46218487394957986, 'f1': 0.43999999999999995, 'number': 119}	{'precision': 0.8016157989228008, 'recall': 0.8384976525821596, 'f1': 0.8196420376319412, 'number': 1065}	0.7519	0.8088	0.7793	0.8255
0.1004	25.0	250	0.8048	{'precision': 0.7424412094064949, 'recall': 0.8195302843016069, 'f1': 0.7790834312573444, 'number': 809}	{'precision': 0.41304347826086957, 'recall': 0.4789915966386555, 'f1': 0.443579766536965, 'number': 119}	{'precision': 0.8048561151079137, 'recall': 0.8403755868544601, 'f1': 0.8222324299494717, 'number': 1065}	0.7536	0.8103	0.7809	0.8256

Framework versions

• Transformers 4.41.2
• Pytorch 2.1.2
• Datasets 2.19.2
• Tokenizers 0.19.1