← Back to LR Schedule

linear warmup

LR Schedule
Used in
8 PRs
Best BPB
1.1896
Avg BPB
1.3630

Hyperparameters Across PRs

pr_numberparameters
46{"warmup_steps":100}
536{"warmup_steps":30}
896{"warmup_steps":200}
1125{"warmup_steps":20}
1163{"warmup_steps":20}
1203{"warmup_steps":5}
1239{"warmup_fraction":0.2}
1691{"warmup_steps":20}