← Back to LR Schedule
adaptive cosine decay
LR ScheduleUsed in
2 PRs
Best BPB
0.2071
Avg BPB
0.6236
Hyperparameters Across PRs
| pr_number | parameters |
|---|---|
| 731 | {"ramp_multiplier_start":1,"ramp_multiplier_end":3,"ramp_fraction":0.3} |
| 851 | {"adaptive_lr":true,"adaptive_lr_max":3} |