Edit model card

layoutlmv3-finetuned-invoice

This model is a fine-tuned version of microsoft/layoutlmv3-base on the generated dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0042
  • Precision: 0.9960
  • Recall: 0.9980
  • F1: 0.9970
  • Accuracy: 0.9996

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: 2
  • eval_batch_size: 2
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • training_steps: 2000

Training results

Training Loss Epoch Step Validation Loss Precision Recall F1 Accuracy
No log 2.0 100 0.1164 0.902 0.9148 0.9084 0.9897
No log 4.0 200 0.0262 0.972 0.9858 0.9789 0.9971
No log 6.0 300 0.0157 0.972 0.9858 0.9789 0.9971
No log 8.0 400 0.0097 0.9877 0.9797 0.9837 0.9979
0.1294 10.0 500 0.0065 0.9939 0.9959 0.9949 0.9994
0.1294 12.0 600 0.0042 0.9960 0.9980 0.9970 0.9996
0.1294 14.0 700 0.0048 0.9960 0.9980 0.9970 0.9996
0.1294 16.0 800 0.0047 0.9960 0.9980 0.9970 0.9996
0.1294 18.0 900 0.0045 0.9960 0.9980 0.9970 0.9996
0.0051 20.0 1000 0.0042 0.9960 0.9980 0.9970 0.9996
0.0051 22.0 1100 0.0041 0.9960 0.9980 0.9970 0.9996
0.0051 24.0 1200 0.0041 0.9960 0.9980 0.9970 0.9996
0.0051 26.0 1300 0.0040 0.9960 0.9980 0.9970 0.9996
0.0051 28.0 1400 0.0040 0.9960 0.9980 0.9970 0.9996
0.0024 30.0 1500 0.0040 0.9960 0.9980 0.9970 0.9996
0.0024 32.0 1600 0.0040 0.9960 0.9980 0.9970 0.9996
0.0024 34.0 1700 0.0039 0.9960 0.9980 0.9970 0.9996
0.0024 36.0 1800 0.0039 0.9960 0.9980 0.9970 0.9996
0.0024 38.0 1900 0.0039 0.9960 0.9980 0.9970 0.9996
0.0018 40.0 2000 0.0039 0.9960 0.9980 0.9970 0.9996

Framework versions

  • Transformers 4.36.0.dev0
  • Pytorch 2.1.0+cu118
  • Datasets 2.14.6
  • Tokenizers 0.14.1
Downloads last month
6
Safetensors
Model size
126M params
Tensor type
F32
·

Finetuned from

Evaluation results