metadata
language:
- en
license: mit
base_model: roberta-large
tags:
- generated_from_trainer
datasets:
- schone-power
model-index:
- name: final
results: []
final
This model is a fine-tuned version of roberta-large on the GLUE SCHONE_POW dataset. It achieves the following results on the evaluation set:
- Loss: 0.1556
- Roc Auc: 0.9742
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: 128
- eval_batch_size: 256
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5.0
Training results
Training Loss | Epoch | Step | Validation Loss | Roc Auc |
---|---|---|---|---|
0.2924 | 0.9985 | 332 | 0.2605 | 0.9248 |
0.2553 | 2.0 | 665 | 0.2234 | 0.9468 |
0.2317 | 2.9985 | 997 | 0.1899 | 0.9620 |
0.2063 | 4.0 | 1330 | 0.1645 | 0.9715 |
0.1897 | 4.9925 | 1660 | 0.1556 | 0.9742 |
Framework versions
- Transformers 4.44.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.20.0
- Tokenizers 0.19.1