← Back to Weight Averaging

LAWA

Weight Averaging
Used in
2 PRs
Best BPB
1.2302
Avg BPB
1.4337

Hyperparameters Across PRs

pr_numberparameters
130{"start_frac":0.8,"checkpoint_interval_steps":200,"avg_last_fraction":0.2}
675{"warmdown_checkpoints":"12-13"}