File size: 2,163 Bytes
f50be2d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 |
## LayoutLMv3-Fine-Tuning-Invoice Model
#### Model description
**LayoutLMv3-Fine-Tuning-Invoice Model** is a fine-tuned version of [microsoft/layoutlmv3-base](https://huggingface.co/microsoft/layoutlmv3-base) on the invoice dataset. For the fine-tuning, We used [Invoice Dataset] that includes 12 labels ('Other', 'ABN', 'BILLER', 'BILLER_ADDRESS', 'BILLER_POST_CODE', 'DUE_DATE', 'GST', 'INVOICE_DATE', 'INVOICE_NUMBER', 'SUBTOTAL', 'TOTAL', 'BILLER_ADDRESS').
This model is a fine-tuned version of [microsoft/layoutlmv3-base](https://huggingface.co/microsoft/layoutlmv3-base) on the invoice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.005334
- Precision: 1.0
- Recall: 1.0
- F1: 1.0
- Accuracy: 1.0
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1.5e-05
- train_batch_size: 2
- eval_batch_size: 2
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- training_steps: 1000
### Training results
| Training Loss | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|:-------------:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
| No log | 100 | 0.0878 | 0.968 | 0.9817 | 0.9748 | 0.9966 |
| No log | 200 | 0.0241 | 0.972 | 0.9858 | 0.9789 | 0.9971 |
| No log | 300 | 0.0186 | 0.972 | 0.9858 | 0.9789 | 0.9971 |
| No log | 400 | 0.0184 | 0.9854 | 0.9574 | 0.9712 | 0.9956 |
| 0.110800 | 500 | 0.0016 | 1.0 | 1.0 | 1.0 | 1.0 |
| 0.110800 | 600 | 0.0015 | 1.0 | 1.0 | 1.0 | 1.0 |
| 0.110800 | 700 | 0.0014 | 1.0 | 1.0 | 1.0 | 1.0 |
| 0.110800 | 800 | 0.0013 | 1.0 | 1.0 | 1.0 | 1.0 |
| 0.110800 | 900 | 0.0012 | 1.0 | 1.0 | 1.0 | 1.0 |
| 0.004900 | 1000 | 0.0012 | 1.0 | 1.0 | 1.0 | 1.0 |
### Framework versions
- Transformers 4.20.0.dev0
- Pytorch 1.11.0+cu113
- Datasets 2.2.2
- Tokenizers 0.12.1 |