icdar23-entrydetector_plaintext
This model is a fine-tuned version of HueyNemud/das22-10-camembert_pretrained on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0424
- Ebegin: {'precision': 0.9725125822686799, 'recall': 0.9447160586686725, 'f1': 0.9584128195345288, 'number': 2659}
- Eend: {'precision': 0.9570211189329382, 'recall': 0.9652466367713004, 'f1': 0.9611162790697675, 'number': 2676}
- Overall Precision: 0.9646
- Overall Recall: 0.9550
- Overall F1: 0.9598
- Overall Accuracy: 0.9923
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: 0.0001
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- training_steps: 7500
Training results
Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
---|---|---|---|---|---|---|---|
No log | 0.07 | 300 | 0.0487 | 0.9874 | 0.9565 | 0.9717 | 0.9943 |
0.1698 | 0.14 | 600 | 0.0310 | 0.9891 | 0.9709 | 0.9799 | 0.9959 |
0.1698 | 0.21 | 900 | 0.0267 | 0.9746 | 0.9764 | 0.9755 | 0.9953 |
0.0346 | 0.29 | 1200 | 0.0217 | 0.9885 | 0.9685 | 0.9784 | 0.9956 |
0.0237 | 0.36 | 1500 | 0.0201 | 0.9866 | 0.9742 | 0.9804 | 0.9960 |
0.0237 | 0.43 | 1800 | 0.0268 | 0.9883 | 0.9561 | 0.9719 | 0.9944 |
0.0205 | 0.5 | 2100 | 0.0216 | 0.9823 | 0.9779 | 0.9801 | 0.9959 |
0.0205 | 0.57 | 2400 | 0.0236 | 0.9874 | 0.9700 | 0.9787 | 0.9957 |
0.0196 | 0.64 | 2700 | 0.0246 | 0.9877 | 0.9668 | 0.9772 | 0.9954 |
0.0195 | 0.72 | 3000 | 0.0254 | 0.9789 | 0.9682 | 0.9735 | 0.9950 |
Framework versions
- Transformers 4.26.1
- Pytorch 1.13.1+cu116
- Datasets 2.9.0
- Tokenizers 0.13.2
- Downloads last month
- 3
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.