invoice_extraction_20240814_layoutlmv3_type_2

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

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