Layoutlm_invoices

This model is a fine-tuned version of microsoft/layoutlm-base-uncased on the layoutlmv4 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0603
• Customer Address: {'precision': 0.7692307692307693, 'recall': 0.9090909090909091, 'f1': 0.8333333333333333, 'number': 11}
• Customer Name: {'precision': 0.8333333333333334, 'recall': 0.9090909090909091, 'f1': 0.8695652173913043, 'number': 11}
• Invoice Number: {'precision': 0.9090909090909091, 'recall': 0.9090909090909091, 'f1': 0.9090909090909091, 'number': 11}
• Tax Amount: {'precision': 1.0, 'recall': 1.0, 'f1': 1.0, 'number': 11}
• Total Amount: {'precision': 1.0, 'recall': 0.9090909090909091, 'f1': 0.9523809523809523, 'number': 11}
• Vendor Name: {'precision': 0.9166666666666666, 'recall': 1.0, 'f1': 0.9565217391304348, 'number': 11}
• Overall Precision: 0.8986
• Overall Recall: 0.9394
• Overall F1: 0.9185
• Overall Accuracy: 0.9831

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: 1e-05
• train_batch_size: 6
• eval_batch_size: 3
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Customer Address	Customer Name	Invoice Number	Tax Amount	Total Amount	Vendor Name	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
0.0714	1.25	10	0.0752	{'precision': 0.9166666666666666, 'recall': 1.0, 'f1': 0.9565217391304348, 'number': 11}	{'precision': 0.9090909090909091, 'recall': 0.9090909090909091, 'f1': 0.9090909090909091, 'number': 11}	{'precision': 0.9090909090909091, 'recall': 0.9090909090909091, 'f1': 0.9090909090909091, 'number': 11}	{'precision': 1.0, 'recall': 1.0, 'f1': 1.0, 'number': 11}	{'precision': 1.0, 'recall': 0.9090909090909091, 'f1': 0.9523809523809523, 'number': 11}	{'precision': 0.9166666666666666, 'recall': 1.0, 'f1': 0.9565217391304348, 'number': 11}	0.9403	0.9545	0.9474	0.9864
0.0572	2.5	20	0.0603	{'precision': 0.7692307692307693, 'recall': 0.9090909090909091, 'f1': 0.8333333333333333, 'number': 11}	{'precision': 0.8333333333333334, 'recall': 0.9090909090909091, 'f1': 0.8695652173913043, 'number': 11}	{'precision': 0.9090909090909091, 'recall': 0.9090909090909091, 'f1': 0.9090909090909091, 'number': 11}	{'precision': 1.0, 'recall': 1.0, 'f1': 1.0, 'number': 11}	{'precision': 1.0, 'recall': 0.9090909090909091, 'f1': 0.9523809523809523, 'number': 11}	{'precision': 0.9166666666666666, 'recall': 1.0, 'f1': 0.9565217391304348, 'number': 11}	0.8986	0.9394	0.9185	0.9831

Framework versions

• Transformers 4.32.1
• Pytorch 2.2.0+cpu
• Datasets 2.12.0
• Tokenizers 0.13.2