gpt-imdb-alpha_0.7-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:
- Step: 7000
- Loss: 11466.6748
- Rewards/chosen: 0.1662
- Rewards/rejected: -0.5317
- Rewards/accuracies: 0.7937
- Rewards/margins: 0.6979
- Logps/rejected: -269.0021
- Logps/chosen: -233.6036
- Logits/rejected: -31.0907
- Logits/chosen: -31.5102
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1.1713 | 0.21 | 500 | 2.5800 | 0.3038 | -0.0770 | 0.7188 | 0.3808 | -264.4555 | -232.2277 | -33.8095 | -34.2861 |
| 0.887 | 0.42 | 1000 | 21.3065 | 0.5505 | 0.1747 | 0.6917 | 0.3758 | -261.9387 | -229.7607 | -32.8001 | -33.3563 |
| 0.798 | 0.63 | 1500 | 61.4252 | 0.5093 | -0.0100 | 0.7333 | 0.5193 | -263.7849 | -230.1718 | -30.8724 | -31.2678 |
| 1.1771 | 0.83 | 2000 | 14.1653 | 0.6467 | 0.1330 | 0.6854 | 0.5138 | -262.3556 | -228.7979 | -33.4203 | -33.7502 |
| 0.5587 | 1.04 | 2500 | 528756.25 | 0.5517 | -0.0428 | 0.7396 | 0.5944 | -264.1129 | -229.7487 | -32.9646 | -33.4291 |
| 0.4833 | 1.25 | 3000 | 1178.0547 | 0.5836 | 0.0507 | 0.6958 | 0.5329 | -263.1786 | -229.4295 | -32.7156 | -33.0784 |
| 0.6214 | 1.46 | 3500 | 4177.1973 | 0.2927 | -0.3473 | 0.7562 | 0.6400 | -267.1580 | -232.3383 | -29.8543 | -30.1578 |
| 18.5015 | 1.67 | 4000 | 513.4760 | 0.4129 | -0.2026 | 0.7479 | 0.6155 | -265.7109 | -231.1364 | -30.7645 | -31.1263 |
| 0.3705 | 1.88 | 4500 | 135.9144 | 0.4609 | -0.1462 | 0.75 | 0.6071 | -265.1470 | -230.6563 | -30.2459 | -30.6495 |
| 0.4778 | 2.08 | 5000 | 1561.6661 | 0.2544 | -0.4144 | 0.7792 | 0.6687 | -267.8289 | -232.7216 | -30.5732 | -30.9863 |
| 0.3125 | 2.29 | 5500 | 8448.3389 | 0.2045 | -0.4842 | 0.7937 | 0.6887 | -268.5275 | -233.2203 | -31.2362 | -31.6616 |
| 6.2284 | 2.5 | 6000 | 13438.1006 | 0.1295 | -0.5751 | 0.7937 | 0.7045 | -269.4362 | -233.9707 | -31.0171 | -31.4348 |
| 2.1427 | 2.71 | 6500 | 13021.2812 | 0.1590 | -0.5409 | 0.7958 | 0.6999 | -269.0947 | -233.6758 | -31.1241 | -31.5456 |
| 24.2387 | 2.92 | 7000 | 11466.6748 | 0.1662 | -0.5317 | 0.7937 | 0.6979 | -269.0021 | -233.6036 | -31.0907 | -31.5102 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.1
- Datasets 2.15.0
- Tokenizers 0.15.0
- Downloads last month
- 6