deberta-toxic-reward-grpo3
This model is a fine-tuned version of microsoft/deberta-v3-base on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.3799
- Accuracy: 0.8056
- Precision: 0.7459
- Recall: 0.9897
- F1: 0.8507
- Auc: 0.8931
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: 64
- eval_batch_size: 16
- seed: 13
- gradient_accumulation_steps: 8
- total_train_batch_size: 512
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 8
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 | Auc |
|---|---|---|---|---|---|---|---|---|
| No log | 1.0 | 71 | 0.3680 | 0.8024 | 0.7432 | 0.9883 | 0.8484 | 0.8832 |
| No log | 2.0 | 142 | 0.3691 | 0.8028 | 0.7442 | 0.9865 | 0.8484 | 0.8877 |
| No log | 3.0 | 213 | 0.4363 | 0.7915 | 0.7288 | 0.9992 | 0.8428 | 0.8768 |
| No log | 4.0 | 284 | 0.3638 | 0.8053 | 0.7451 | 0.9909 | 0.8506 | 0.8914 |
| No log | 5.0 | 355 | 0.3637 | 0.8045 | 0.7467 | 0.9845 | 0.8493 | 0.8920 |
| No log | 6.0 | 426 | 0.3801 | 0.8056 | 0.7459 | 0.9897 | 0.8507 | 0.8931 |
| No log | 7.0 | 497 | 0.3841 | 0.8045 | 0.7467 | 0.9845 | 0.8493 | 0.8931 |
| 0.3585 | 8.0 | 568 | 0.3898 | 0.8046 | 0.7483 | 0.9807 | 0.8489 | 0.8927 |
Framework versions
- Transformers 4.57.1
- Pytorch 2.9.1+cu130
- Datasets 4.4.1
- Tokenizers 0.22.1