layoutlmv3-finetuned-cord
This model is a fine-tuned version of microsoft/layoutlmv3-base on the CORD dataset. It achieves the following results on the evaluation set:
- Loss: 0.1845
- Precision: 0.9620
- Recall: 0.9656
- F1: 0.9638
- Accuracy: 0.9682
The script for training can be found here: https://github.com/huggingface/transformers/tree/main/examples/research_projects/layoutlmv3
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: 5e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| No log | 2.0 | 100 | 0.5257 | 0.8223 | 0.8555 | 0.8386 | 0.8710 |
| No log | 4.0 | 200 | 0.3200 | 0.9118 | 0.9281 | 0.9199 | 0.9317 |
| No log | 6.0 | 300 | 0.2449 | 0.9298 | 0.9424 | 0.9361 | 0.9465 |
| No log | 8.0 | 400 | 0.1923 | 0.9472 | 0.9536 | 0.9504 | 0.9597 |
| 0.4328 | 10.0 | 500 | 0.1857 | 0.9591 | 0.9656 | 0.9623 | 0.9682 |
| 0.4328 | 12.0 | 600 | 0.2073 | 0.9597 | 0.9618 | 0.9607 | 0.9656 |
| 0.4328 | 14.0 | 700 | 0.1804 | 0.9634 | 0.9663 | 0.9649 | 0.9703 |
| 0.4328 | 16.0 | 800 | 0.1882 | 0.9634 | 0.9648 | 0.9641 | 0.9665 |
| 0.4328 | 18.0 | 900 | 0.1800 | 0.9619 | 0.9648 | 0.9634 | 0.9677 |
| 0.0318 | 20.0 | 1000 | 0.1845 | 0.9620 | 0.9656 | 0.9638 | 0.9682 |
Framework versions
- Transformers 4.19.0.dev0
- Pytorch 1.11.0+cu113
- Datasets 2.0.0
- Tokenizers 0.11.6