iw-conspir-v3
This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.3976
- F1 Contradictory: 0.0
- F1 Overriding Suspicion: 0.7035
- F1 Nefarious Intent: 0.6841
- F1 Persecuted Victim: 0.2857
- F1 Immune To Evidence: 0.3396
- F1 Reinterpreting Randomness: 0.4010
- Macro F1: 0.4023
- Micro F1: 0.5893
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: 16
- eval_batch_size: 32
- seed: 42
- 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
- lr_scheduler_warmup_steps: 0.1
- num_epochs: 6
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | F1 Contradictory | F1 Overriding Suspicion | F1 Nefarious Intent | F1 Persecuted Victim | F1 Immune To Evidence | F1 Reinterpreting Randomness | Macro F1 | Micro F1 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.2721 | 1.0 | 100 | 0.3105 | 0.0 | 0.6 | 0.5649 | 0.2092 | 0.28 | 0.3920 | 0.3410 | 0.4886 |
| 0.2332 | 2.0 | 200 | 0.3372 | 0.0177 | 0.6601 | 0.6708 | 0.3167 | 0.3344 | 0.3814 | 0.3968 | 0.5335 |
| 0.1490 | 3.0 | 300 | 0.3975 | 0.0 | 0.7035 | 0.6841 | 0.2857 | 0.3396 | 0.4010 | 0.4023 | 0.5893 |
| 0.1420 | 4.0 | 400 | 0.4999 | 0.0 | 0.7089 | 0.6701 | 0.2532 | 0.3314 | 0.3830 | 0.3911 | 0.5987 |
| 0.1038 | 5.0 | 500 | 0.5314 | 0.0 | 0.6874 | 0.6601 | 0.2532 | 0.3541 | 0.4070 | 0.3936 | 0.5871 |
| 0.0963 | 6.0 | 600 | 0.5686 | 0.0 | 0.6617 | 0.6628 | 0.2133 | 0.3263 | 0.3878 | 0.3753 | 0.5741 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.0.0
- Tokenizers 0.22.2