lr = 5e-05per_device_batch_size = 8gradient_accumulation_steps = 2weight_decay = 1e-09seed = 42numerical stability denominator constant = 0.001lambda = 0.01alpha = 2.0beta = 2.0| eval_loss | eval_accuracy | epoch |
|---|---|---|
| 19.394 | 0.208 | 0.67 |
| 13.088 | 0.792 | 2.0 |
| 12.683 | 0.792 | 2.67 |
| 12.690 | 0.792 | 4.0 |
| 13.599 | 0.917 | 4.67 |
| 14.317 | 0.708 | 6.0 |
| 13.274 | 0.750 | 6.67 |
| 10.404 | 0.917 | 8.0 |
| 10.809 | 0.917 | 8.67 |
| 12.907 | 0.833 | 10.0 |
| 14.400 | 0.833 | 10.67 |
| 12.761 | 0.833 | 12.0 |
| 12.600 | 0.833 | 12.67 |