albert-large-v2-finetuned-ner_with_callbacks
This model is a fine-tuned version of albert-large-v2 on the PLOD-unfiltered dataset. It achieves the following results on the evaluation set:
- Loss: 0.1235
- Precision: 0.9655
- Recall: 0.9608
- F1: 0.9632
- Accuracy: 0.9589
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: 8
- eval_batch_size: 4
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 6
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.1377 | 0.49 | 7000 | 0.1294 | 0.9563 | 0.9422 | 0.9492 | 0.9436 |
| 0.1244 | 0.98 | 14000 | 0.1165 | 0.9589 | 0.9504 | 0.9546 | 0.9499 |
| 0.107 | 1.48 | 21000 | 0.1140 | 0.9603 | 0.9509 | 0.9556 | 0.9511 |
| 0.1088 | 1.97 | 28000 | 0.1086 | 0.9613 | 0.9551 | 0.9582 | 0.9536 |
| 0.0918 | 2.46 | 35000 | 0.1059 | 0.9617 | 0.9582 | 0.9600 | 0.9556 |
| 0.0847 | 2.95 | 42000 | 0.1067 | 0.9620 | 0.9586 | 0.9603 | 0.9559 |
| 0.0734 | 3.44 | 49000 | 0.1188 | 0.9646 | 0.9588 | 0.9617 | 0.9574 |
| 0.0725 | 3.93 | 56000 | 0.1065 | 0.9660 | 0.9599 | 0.9630 | 0.9588 |
| 0.0547 | 4.43 | 63000 | 0.1273 | 0.9662 | 0.9602 | 0.9632 | 0.9590 |
| 0.0542 | 4.92 | 70000 | 0.1235 | 0.9655 | 0.9608 | 0.9632 | 0.9589 |
| 0.0374 | 5.41 | 77000 | 0.1401 | 0.9647 | 0.9613 | 0.9630 | 0.9586 |
| 0.0417 | 5.9 | 84000 | 0.1380 | 0.9641 | 0.9622 | 0.9632 | 0.9588 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.10.1+cu111
- Datasets 2.1.0
- Tokenizers 0.12.1