Add sageattn mode that allows torch.compile

Latest wheel from woct0rdho includes the torch.compile fix: https://github.com/woct0rdho/SageAttention/releases Based on my quick testing this reduces peak VRAM usage a bit when running sageattn + torch.compile
2025-10-20 15:16:43 +03:00
parent 8081e1337c
commit 200f6943e3
2 changed files with 11 additions and 0 deletions
@@ -1002,6 +1002,7 @@ class WanVideoModelLoader:
                    "sageattn",
                    "sageattn_3",
                    "radial_sage_attention",
+                    "sageattn_compiled",
                    ], {"default": "sdpa"}),
                "compile_args": ("WANCOMPILEARGS", ),
                "block_swap_args": ("BLOCKSWAPARGS", ),