← Back to Architecture

SparseAttnGate

Architecture
Used in
10 PRs
Best BPB
0.9019
Avg BPB
1.0460

Hyperparameters Across PRs

pr_numberparameters
1869—
1881—
1883—
1915—
1935—
1945—
2009—
2014{"gate_window":12,"scale":0.5}
2051—
2143{"params_per_layer":96}