← Back to Architecture
Partial RoPE + NTK-aware scaling
ArchitectureUsed in
1 PRs
Best BPB
1.1175
Avg BPB
1.1175
Submissions
Hyperparameters Across PRs
| pr_number | parameters |
|---|---|
| 569 | {"partial_dims":[16,64],"ntk_base":10000} |