| === Training Metrics Report === | |
| Current State: | |
| Step: 5000 | |
| Tokens Processed: 819,200,000 | |
| Training Progress: 305.18% | |
| Recent Metrics: | |
| Step 4000: | |
| loss: 1.7636 | |
| grad_norm: 5.131964206695557 | |
| learning_rate: 0.00028000000000000003 | |
| epoch: 0.2 | |
| Step 4250: | |
| loss: 1.6028 | |
| grad_norm: 4.24104118347168 | |
| learning_rate: 0.00029749999999999997 | |
| epoch: 0.2125 | |
| Step 4500: | |
| loss: 1.5 | |
| grad_norm: 3.984833240509033 | |
| learning_rate: 0.000315 | |
| epoch: 0.225 | |
| Step 4750: | |
| loss: 1.425 | |
| grad_norm: 3.9891703128814697 | |
| learning_rate: 0.0003325 | |
| epoch: 0.2375 | |
| Step 5000: | |
| loss: 1.3692 | |
| grad_norm: 3.614696502685547 | |
| learning_rate: 0.00035 | |
| epoch: 0.25 | |