mdeberta-hate-final
This model is a fine-tuned version of microsoft/mdeberta-v3-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.6223
- Accuracy: 0.7424
- Precision: 0.7410
- Recall: 0.7424
- F1: 0.7363
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: 1e-05
- train_batch_size: 32
- eval_batch_size: 32
- 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 | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 296 | 0.5309 | 0.7519 | 0.7685 | 0.7519 | 0.7357 |
| 0.5358 | 2.0 | 592 | 0.5228 | 0.7510 | 0.7663 | 0.7510 | 0.7351 |
| 0.5358 | 3.0 | 888 | 0.5565 | 0.7510 | 0.7513 | 0.7510 | 0.7438 |
| 0.4295 | 4.0 | 1184 | 0.5639 | 0.7481 | 0.7488 | 0.7481 | 0.7403 |
| 0.4295 | 5.0 | 1480 | 0.5941 | 0.7510 | 0.7531 | 0.7510 | 0.7423 |
| 0.3701 | 6.0 | 1776 | 0.6223 | 0.7424 | 0.7410 | 0.7424 | 0.7363 |
Framework versions
- Transformers 4.24.0.dev0
- Pytorch 1.11.0+cu102
- Datasets 2.6.1
- Tokenizers 0.13.1