gpt-imdb-alpha_0.5-beta_0.1
This model is a fine-tuned version of lvwerra/gpt2-imdb on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 5886.0698
- Rewards/chosen: -0.0051
- Rewards/rejected: -0.7543
- Rewards/accuracies: 0.8125
- Rewards/margins: 0.7492
- Logps/rejected: -271.2288
- Logps/chosen: -235.3164
- Logits/rejected: -35.8752
- Logits/chosen: -36.3770
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: 24
- eval_batch_size: 24
- seed: 42
- optimizer: Adam with betas=(0.9,0.99) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 150
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.4827 | 0.21 | 500 | 1.0040 | 0.1562 | -0.2070 | 0.7042 | 0.3632 | -265.7552 | -233.7028 | -33.0609 | -33.6065 |
| 14.1335 | 0.42 | 1000 | 3.0758 | 0.3762 | -0.1852 | 0.7438 | 0.5615 | -265.5375 | -231.5030 | -34.4930 | -35.0582 |
| 0.5469 | 0.63 | 1500 | 8.0814 | 0.4345 | -0.1070 | 0.7271 | 0.5415 | -264.7556 | -230.9207 | -32.8344 | -33.3794 |
| 1.032 | 0.83 | 2000 | 4.5711 | 0.5518 | -0.0259 | 0.7104 | 0.5777 | -263.9442 | -229.7469 | -33.4042 | -33.9772 |
| 0.3719 | 1.04 | 2500 | 459.9075 | 0.2914 | -0.4286 | 0.7813 | 0.7200 | -267.9716 | -232.3516 | -33.0798 | -33.6079 |
| 0.4085 | 1.25 | 3000 | 526.3080 | 0.4340 | -0.2325 | 0.7479 | 0.6666 | -266.0108 | -230.9248 | -35.2424 | -35.7675 |
| 2.1291 | 1.46 | 3500 | 630.5800 | 0.4165 | -0.3073 | 0.7604 | 0.7238 | -266.7581 | -231.0998 | -37.0077 | -37.6012 |
| 4.7118 | 1.67 | 4000 | 96.2745 | 0.3115 | -0.4479 | 0.7625 | 0.7593 | -268.1639 | -232.1506 | -37.1158 | -37.6120 |
| 0.5195 | 1.88 | 4500 | 342.8383 | 0.3188 | -0.4079 | 0.7688 | 0.7267 | -267.7646 | -232.0775 | -37.3006 | -37.8729 |
| 0.8474 | 2.08 | 5000 | 4552.9634 | -0.0725 | -0.8330 | 0.7896 | 0.7605 | -272.0149 | -235.9899 | -36.5234 | -37.0376 |
| 0.2874 | 2.29 | 5500 | 3540.6086 | 0.0246 | -0.7477 | 0.8083 | 0.7723 | -271.1625 | -235.0193 | -36.0173 | -36.5541 |
| 2.4701 | 2.5 | 6000 | 4522.3066 | -0.0217 | -0.7825 | 0.8042 | 0.7608 | -271.5105 | -235.4827 | -35.7649 | -36.2731 |
| 0.59 | 2.71 | 6500 | 4948.8481 | 0.0070 | -0.7472 | 0.8104 | 0.7542 | -271.1574 | -235.1950 | -35.8213 | -36.3258 |
| 0.3244 | 2.92 | 7000 | 5886.0698 | -0.0051 | -0.7543 | 0.8125 | 0.7492 | -271.2288 | -235.3164 | -35.8752 | -36.3770 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.1
- Datasets 2.15.0
- Tokenizers 0.15.0
- Downloads last month
- 1