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: 1.5594
• Answer: {'precision': 0.03052064631956912, 'recall': 0.042027194066749075, 'f1': 0.035361414456578255, 'number': 809}
• Header: {'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}
• Question: {'precision': 0.18327402135231316, 'recall': 0.19342723004694837, 'f1': 0.18821379625399726, 'number': 1065}
• Overall Precision: 0.1072
• Overall Recall: 0.1204
• Overall F1: 0.1134
• Overall Accuracy: 0.4038

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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Answer	Header	Question	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
1.8318	1.0	10	1.6681	{'precision': 0.009746588693957114, 'recall': 0.012360939431396786, 'f1': 0.010899182561307902, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.0981169474727453, 'recall': 0.09295774647887324, 'f1': 0.09546769527483126, 'number': 1065}	0.0535	0.0547	0.0541	0.3444
1.5836	2.0	20	1.5594	{'precision': 0.03052064631956912, 'recall': 0.042027194066749075, 'f1': 0.035361414456578255, 'number': 809}	{'precision': 0.0, 'recall': 0.0, 'f1': 0.0, 'number': 119}	{'precision': 0.18327402135231316, 'recall': 0.19342723004694837, 'f1': 0.18821379625399726, 'number': 1065}	0.1072	0.1204	0.1134	0.4038

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0