← Back to LR Schedule

warmup + warmdown

LR Schedule
Used in
12 PRs
Best BPB
0.6364
Avg BPB
1.1288

Hyperparameters Across PRs

pr_numberparameters
160{"warmup_steps":20,"warmdown_iters":3000}
164{"warmup_steps":1500,"warmdown_steps":3000}
254{"warmup_steps":1500,"warmdown_steps":3000}
256{"warmup_steps":20,"warmdown_iters":3000}
287{"warmup_steps":1500,"warmdown_steps":3000}
305{"warmup_steps":20,"warmdown_iters":3000}
400{"warmup_steps":20,"warmdown_iters":3000}
451{"warmup_steps":20,"warmdown_iters":3000}
665{"warmup_steps":20,"warmdown_iters":3000}
746{"warmup_steps":20,"warmdown_iters":1200}
808{"warmup_steps":1500,"warmdown_iters":3500}
858{"warmup_steps":20,"warmdown_iterations":1200}