<!-- 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. -->
EElayoutlmv3_jordyvl_rvl_cdip_100_examples_per_class_2023-10-29_two_text_visual
This model is a fine-tuned version of jordyvl/LayoutLMv3_RVL-CDIP_NK100 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.0358
- Accuracy: 0.795
- Exit 0 Accuracy: 0.1925
- Exit 1 Accuracy: 0.0725
- Exit 2 Accuracy: 0.0625
- Exit 3 Accuracy: 0.0875
- Exit 4 Accuracy: 0.775
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: 2e-05
- train_batch_size: 2
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 24
- total_train_batch_size: 48
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 60
Training results
Training Loss | Epoch | Step | Validation Loss | Accuracy | Exit 0 Accuracy | Exit 1 Accuracy | Exit 2 Accuracy | Exit 3 Accuracy | Exit 4 Accuracy |
---|---|---|---|---|---|---|---|---|---|
No log | 0.96 | 16 | 0.9549 | 0.7975 | 0.07 | 0.0625 | 0.0625 | 0.0775 | 0.3425 |
No log | 1.98 | 33 | 0.9581 | 0.7975 | 0.0725 | 0.0625 | 0.0625 | 0.0625 | 0.3475 |
No log | 3.0 | 50 | 0.9613 | 0.7975 | 0.09 | 0.0625 | 0.0625 | 0.0625 | 0.395 |
No log | 3.96 | 66 | 0.9641 | 0.7975 | 0.09 | 0.0625 | 0.0625 | 0.08 | 0.44 |
No log | 4.98 | 83 | 0.9669 | 0.7975 | 0.0925 | 0.0625 | 0.0625 | 0.0625 | 0.45 |
No log | 6.0 | 100 | 0.9694 | 0.7975 | 0.1 | 0.0625 | 0.0625 | 0.0625 | 0.495 |
No log | 6.96 | 116 | 0.9716 | 0.7975 | 0.0975 | 0.0625 | 0.0625 | 0.0625 | 0.51 |
No log | 7.98 | 133 | 0.9741 | 0.7975 | 0.0975 | 0.0625 | 0.0625 | 0.0625 | 0.515 |
No log | 9.0 | 150 | 0.9767 | 0.7975 | 0.1 | 0.0625 | 0.0625 | 0.0625 | 0.53 |
No log | 9.96 | 166 | 0.9789 | 0.7975 | 0.1025 | 0.0625 | 0.0625 | 0.0625 | 0.575 |
No log | 10.98 | 183 | 0.9815 | 0.7975 | 0.1125 | 0.0625 | 0.0625 | 0.0625 | 0.5825 |
No log | 12.0 | 200 | 0.9843 | 0.7975 | 0.12 | 0.0625 | 0.0625 | 0.0625 | 0.6175 |
No log | 12.96 | 216 | 0.9865 | 0.7975 | 0.1225 | 0.0625 | 0.0625 | 0.0625 | 0.635 |
No log | 13.98 | 233 | 0.9887 | 0.7975 | 0.12 | 0.0625 | 0.0625 | 0.0625 | 0.66 |
No log | 15.0 | 250 | 0.9909 | 0.7975 | 0.1225 | 0.0625 | 0.0625 | 0.0625 | 0.6775 |
No log | 15.96 | 266 | 0.9933 | 0.7975 | 0.135 | 0.0625 | 0.0625 | 0.0625 | 0.6925 |
No log | 16.98 | 283 | 0.9955 | 0.7975 | 0.14 | 0.0625 | 0.0625 | 0.0625 | 0.7125 |
No log | 18.0 | 300 | 0.9979 | 0.7975 | 0.135 | 0.0625 | 0.0625 | 0.1 | 0.7075 |
No log | 18.96 | 316 | 0.9999 | 0.7975 | 0.1275 | 0.0625 | 0.0625 | 0.1 | 0.7075 |
No log | 19.98 | 333 | 1.0017 | 0.7975 | 0.1275 | 0.0625 | 0.0625 | 0.0675 | 0.7225 |
No log | 21.0 | 350 | 1.0033 | 0.7975 | 0.125 | 0.0625 | 0.0625 | 0.0625 | 0.735 |
No log | 21.96 | 366 | 1.0047 | 0.7975 | 0.1275 | 0.0625 | 0.0625 | 0.0625 | 0.765 |
No log | 22.98 | 383 | 1.0065 | 0.7975 | 0.1325 | 0.0625 | 0.0625 | 0.0625 | 0.76 |
No log | 24.0 | 400 | 1.0082 | 0.7975 | 0.1275 | 0.095 | 0.065 | 0.0625 | 0.7675 |
No log | 24.96 | 416 | 1.0099 | 0.7975 | 0.1325 | 0.06 | 0.1 | 0.0625 | 0.78 |
No log | 25.98 | 433 | 1.0114 | 0.7975 | 0.13 | 0.06 | 0.0975 | 0.0625 | 0.7825 |
No log | 27.0 | 450 | 1.0129 | 0.7975 | 0.13 | 0.0725 | 0.09 | 0.0625 | 0.79 |
No log | 27.96 | 466 | 1.0143 | 0.795 | 0.135 | 0.105 | 0.0625 | 0.0625 | 0.7825 |
No log | 28.98 | 483 | 1.0158 | 0.795 | 0.14 | 0.095 | 0.0625 | 0.0625 | 0.775 |
2.5042 | 30.0 | 500 | 1.0174 | 0.795 | 0.15 | 0.115 | 0.0625 | 0.0625 | 0.7775 |
2.5042 | 30.96 | 516 | 1.0187 | 0.795 | 0.1425 | 0.095 | 0.0625 | 0.0625 | 0.7775 |
2.5042 | 31.98 | 533 | 1.0200 | 0.795 | 0.1425 | 0.1175 | 0.0625 | 0.0625 | 0.7775 |
2.5042 | 33.0 | 550 | 1.0214 | 0.795 | 0.1525 | 0.09 | 0.0625 | 0.0625 | 0.7725 |
2.5042 | 33.96 | 566 | 1.0227 | 0.795 | 0.1675 | 0.08 | 0.0625 | 0.0625 | 0.77 |
2.5042 | 34.98 | 583 | 1.0238 | 0.795 | 0.175 | 0.085 | 0.0625 | 0.085 | 0.77 |
2.5042 | 36.0 | 600 | 1.0248 | 0.795 | 0.1775 | 0.0775 | 0.0625 | 0.105 | 0.7725 |
2.5042 | 36.96 | 616 | 1.0259 | 0.795 | 0.1825 | 0.085 | 0.0625 | 0.105 | 0.7675 |
2.5042 | 37.98 | 633 | 1.0270 | 0.795 | 0.1975 | 0.08 | 0.0625 | 0.125 | 0.76 |
2.5042 | 39.0 | 650 | 1.0278 | 0.795 | 0.2 | 0.08 | 0.0625 | 0.11 | 0.765 |
2.5042 | 39.96 | 666 | 1.0285 | 0.795 | 0.1875 | 0.07 | 0.0625 | 0.11 | 0.7625 |
2.5042 | 40.98 | 683 | 1.0292 | 0.795 | 0.18 | 0.07 | 0.0625 | 0.115 | 0.7625 |
2.5042 | 42.0 | 700 | 1.0301 | 0.795 | 0.1875 | 0.065 | 0.0625 | 0.115 | 0.7625 |
2.5042 | 42.96 | 716 | 1.0308 | 0.795 | 0.19 | 0.0625 | 0.0625 | 0.115 | 0.76 |
2.5042 | 43.98 | 733 | 1.0314 | 0.795 | 0.1975 | 0.0625 | 0.0625 | 0.1075 | 0.7625 |
2.5042 | 45.0 | 750 | 1.0320 | 0.795 | 0.2 | 0.0625 | 0.0625 | 0.1075 | 0.7675 |
2.5042 | 45.96 | 766 | 1.0325 | 0.795 | 0.2025 | 0.065 | 0.0625 | 0.1075 | 0.7675 |
2.5042 | 46.98 | 783 | 1.0331 | 0.795 | 0.205 | 0.075 | 0.0625 | 0.105 | 0.7675 |
2.5042 | 48.0 | 800 | 1.0336 | 0.795 | 0.2 | 0.075 | 0.0625 | 0.1025 | 0.77 |
2.5042 | 48.96 | 816 | 1.0341 | 0.795 | 0.2025 | 0.08 | 0.0625 | 0.1025 | 0.775 |
2.5042 | 49.98 | 833 | 1.0345 | 0.795 | 0.1825 | 0.0825 | 0.0625 | 0.0975 | 0.775 |
2.5042 | 51.0 | 850 | 1.0348 | 0.795 | 0.1825 | 0.0825 | 0.0625 | 0.1 | 0.775 |
2.5042 | 51.96 | 866 | 1.0351 | 0.795 | 0.1825 | 0.0775 | 0.0625 | 0.105 | 0.775 |
2.5042 | 52.98 | 883 | 1.0353 | 0.795 | 0.185 | 0.0775 | 0.0625 | 0.0975 | 0.775 |
2.5042 | 54.0 | 900 | 1.0355 | 0.795 | 0.1875 | 0.0825 | 0.0625 | 0.09 | 0.775 |
2.5042 | 54.96 | 916 | 1.0357 | 0.795 | 0.195 | 0.0775 | 0.0625 | 0.0875 | 0.775 |
2.5042 | 55.98 | 933 | 1.0357 | 0.795 | 0.1925 | 0.0725 | 0.0625 | 0.09 | 0.775 |
2.5042 | 57.0 | 950 | 1.0358 | 0.795 | 0.1925 | 0.0725 | 0.0625 | 0.09 | 0.775 |
2.5042 | 57.6 | 960 | 1.0358 | 0.795 | 0.1925 | 0.0725 | 0.0625 | 0.0875 | 0.775 |
Framework versions
- Transformers 4.31.0
- Pytorch 2.0.1+cu117
- Datasets 2.13.1
- Tokenizers 0.13.3