← Back to Architecture

QK-Gain

Architecture
Used in
11 PRs
Best BPB
0.9354
Avg BPB
1.0659

Hyperparameters Across PRs

pr_numberparameters
1236{"init":4}
1263{"init":4}
1303{"version":4}
1334{"gain":5}
1364{"gain":4}
1392{"gain":5}
1395{"value":5}
1485{"gain":5}
1512{"gain":2.5}
1532{"gain":5.25}
1731{"gain":5.25}