invoice_extraction_20240814_layoutlmv3_type_2
This model is a fine-tuned version of
microsoft/layoutlmv3-base.
Model description
Trained from LayoutLMv3 base model (microsoft/layoutlmv3-base) with label-studio manually labeled type-2 invoice data
Intended uses & limitations
Parse Traditional Chinese invoices and extract invoice number, date, untaxed price, tax, and total price.
Training and evaluation data
Train data: 25 samples of type 2 label-studio manually labeled images.
Evaluation data: 12 samples of type 2 label-studio manually labeled images.
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- lr_scheduler_type: linear
- num_epochs: 28 (Trained for 40 epochs yet the model with the best f1 score is at Epoch 28)
- metric_for_best_model: "f1"
Training results
TrainOutput(global_step=1000, training_loss=0.16577241975069046, metrics={'train_runtime': 404.5902, 'train_samples_per_second': 2.472, 'train_steps_per_second': 2.472, 'total_flos': 263600618496000.0, 'train_loss': 0.16577241975069046, 'epoch': 40.0}
Framework versions
- Transformers 4.38.2
- Pytorch 2.2.1+cu121
- Datasets 2.14.5
- Tokenizers 0.15.2
