Edit model card

layoutxlm-finetuned-kumon

This model is a fine-tuned version of microsoft/layoutxlm-base on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0869
  • γ‚«γƒƒγƒˆε―Έζ³• Precision: 0.8571
  • γ‚«γƒƒγƒˆε―Έζ³• Recall: 0.5455
  • γ‚«γƒƒγƒˆε―Έζ³• F1: 0.6667
  • γ‚«γƒƒγƒˆε―Έζ³• Number: 22
  • γƒ•γƒ«γƒΌγƒˆ Precision: 0.8947
  • γƒ•γƒ«γƒΌγƒˆ Recall: 0.9444
  • γƒ•γƒ«γƒΌγƒˆ F1: 0.9189
  • γƒ•γƒ«γƒΌγƒˆ Number: 18
  • 寸法 Precision: 0.8
  • 寸法 Recall: 1.0
  • 寸法 F1: 0.8889
  • 寸法 Number: 16
  • 数量 Precision: 0.95
  • 数量 Recall: 0.95
  • 数量 F1: 0.9500
  • 数量 Number: 20
  • 材θ³ͺ Precision: 1.0
  • 材θ³ͺ Recall: 1.0
  • 材θ³ͺ F1: 1.0
  • 材θ³ͺ Number: 24
  • 納期 Precision: 1.0
  • 納期 Recall: 1.0
  • 納期 F1: 1.0
  • 納期 Number: 21
  • Overall Precision: 0.9237
  • Overall Recall: 0.9008
  • Overall F1: 0.9121
  • Overall Accuracy: 0.9862

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: 2e-05
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 0
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 10
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss γ‚«γƒƒγƒˆε―Έζ³• Precision γ‚«γƒƒγƒˆε―Έζ³• Recall γ‚«γƒƒγƒˆε―Έζ³• F1 γ‚«γƒƒγƒˆε―Έζ³• Number γƒ•γƒ«γƒΌγƒˆ Precision γƒ•γƒ«γƒΌγƒˆ Recall γƒ•γƒ«γƒΌγƒˆ F1 γƒ•γƒ«γƒΌγƒˆ Number 寸法 Precision 寸法 Recall 寸法 F1 寸法 Number 数量 Precision 数量 Recall 数量 F1 数量 Number 材θ³ͺ Precision 材θ³ͺ Recall 材θ³ͺ F1 材θ³ͺ Number 納期 Precision 納期 Recall 納期 F1 納期 Number Overall Precision Overall Recall Overall F1 Overall Accuracy
0.3146 1.0 23 0.3481 0.0 0.0 0.0 22 0.0 0.0 0.0 18 0.0 0.0 0.0 16 0.0 0.0 0.0 20 0.0 0.0 0.0 24 0.0 0.0 0.0 21 0.0 0.0 0.0 0.8930
0.3295 2.0 46 0.1314 0.4762 0.4545 0.4651 22 0.9167 0.6111 0.7333 18 0.6522 0.9375 0.7692 16 0.9474 0.9 0.9231 20 0.6087 0.5833 0.5957 24 0.5769 0.7143 0.6383 21 0.6694 0.6860 0.6776 0.9641
0.1176 3.0 69 0.0860 0.7857 0.5 0.6111 22 0.8889 0.8889 0.8889 18 0.8 1.0 0.8889 16 1.0 0.95 0.9744 20 1.0 0.8333 0.9091 24 1.0 0.9048 0.9500 21 0.9182 0.8347 0.8745 0.9805
0.0126 4.0 92 0.0756 0.8571 0.5455 0.6667 22 0.8947 0.9444 0.9189 18 0.8 1.0 0.8889 16 0.95 0.95 0.9500 20 0.9524 0.8333 0.8889 24 0.9524 0.9524 0.9524 21 0.9043 0.8595 0.8814 0.9817
0.0071 5.0 115 0.0743 0.8571 0.5455 0.6667 22 0.8947 0.9444 0.9189 18 0.8 1.0 0.8889 16 0.95 0.95 0.9500 20 1.0 0.9167 0.9565 24 1.0 1.0 1.0 21 0.9224 0.8843 0.9030 0.9849
0.0044 6.0 138 0.0849 0.8571 0.5455 0.6667 22 0.8947 0.9444 0.9189 18 0.8 1.0 0.8889 16 0.95 0.95 0.9500 20 1.0 0.9583 0.9787 24 1.0 1.0 1.0 21 0.9231 0.8926 0.9076 0.9855
0.0018 7.0 161 0.0796 0.8571 0.5455 0.6667 22 0.8947 0.9444 0.9189 18 0.8 1.0 0.8889 16 0.95 0.95 0.9500 20 1.0 0.9583 0.9787 24 1.0 1.0 1.0 21 0.9231 0.8926 0.9076 0.9855
0.0017 8.0 184 0.0836 0.8571 0.5455 0.6667 22 0.8947 0.9444 0.9189 18 0.8 1.0 0.8889 16 0.95 0.95 0.9500 20 1.0 1.0 1.0 24 1.0 1.0 1.0 21 0.9237 0.9008 0.9121 0.9862
0.0032 9.0 207 0.0869 0.8571 0.5455 0.6667 22 0.8947 0.9444 0.9189 18 0.8 1.0 0.8889 16 0.95 0.95 0.9500 20 1.0 1.0 1.0 24 1.0 1.0 1.0 21 0.9237 0.9008 0.9121 0.9862
0.0048 10.0 230 0.0869 0.8571 0.5455 0.6667 22 0.8947 0.9444 0.9189 18 0.8 1.0 0.8889 16 0.95 0.95 0.9500 20 1.0 1.0 1.0 24 1.0 1.0 1.0 21 0.9237 0.9008 0.9121 0.9862

Framework versions

  • Transformers 4.38.2
  • Pytorch 2.1.0+cu121
  • Datasets 2.18.0
  • Tokenizers 0.15.2
Downloads last month
223
Safetensors
Model size
369M params
Tensor type
F32
Β·

Finetuned from