[Perf] Keep non-speculative GDN decode on the standard path (#59735)

Signed-off-by: DimensionSTP <65501090+DimensionSTP@users.noreply.github.com>
Signed-off-by: Thien Tran <gau.nernst@yahoo.com.sg>
Co-authored-by: Thien Tran <gau.nernst@yahoo.com.sg>
Co-authored-by: Kimi Code <noreply@moonshot.ai>
This commit is contained in:
DimensionSTP
2026-10-02 13:44:53 +00:00
committed by GitHub
co-authored by Thien Tran Kimi Code
parent 24df2d3fc7
commit e4294144a8
2 changed files with 14 additions and 9 deletions
+1 -1
View File
@@ -194,7 +194,7 @@ def test_fused_forward_uses_packed_entrypoint() -> None:
ba = torch.randn(num_tokens, 2 * HV, dtype=torch.bfloat16, device=device)
layer = types.SimpleNamespace(
prefix=PREFIX,
enable_fused_gdn_decode=True,
enable_fused_gdn_spec_decode=True,
norm=types.SimpleNamespace(
weight=torch.empty(V, dtype=torch.bfloat16, device=device)
),