← Back to Architecture

Sparse Attention Gate

Architecture
Used in
3 PRs
Best BPB
1.0586
Avg BPB
1.0647

Submissions

PR #1855by codemath3000RECORD
1.0611
PR #1953by andrewbaggio1RECORD
1.0586
PR #2088by MaxIv25
1.0744

Hyperparameters Across PRs

pr_numberparameters
1855{"gate_window":12,"scale":0.5}
1953—
2088—