[ROCm] gfx1250 optimizations and model enablement (#58800)

Signed-off-by: jpvillam <juan.villamizar@amd.com>
Signed-off-by: Jaden Mathias <jaden.mathias@amd.com>
Signed-off-by: Juan Villamizar <jpvillam@amd.com>
Signed-off-by: jpvillam <jpvillam@amd.com>
Signed-off-by: Omar Muhammad <omar.muhammad@amd.com>
Signed-off-by: Daniel <danichan@amd.com>
Signed-off-by: Juan Villamizar <100237675+jpvillam-amd@users.noreply.github.com>
Co-authored-by: Jaden Mathias <jaden.mathias@amd.com>
Co-authored-by: jpvillam <jpvillam@amd.com>
Co-authored-by: Omar Muhammad <omar.muhammad@amd.com>
Co-authored-by: Daniel <danichan@amd.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Daniel Chan <Daniel.Chan@amd.com>
This commit is contained in:
Juan Villamizar
2026-10-15 14:51:57 -07:00
committed by GitHub
co-authored by Jaden Mathias jpvillam Omar Muhammad Daniel Claude Opus 5 Daniel Chan
parent 9b99ccb56b
commit e711723352
21 changed files with 845 additions and 73 deletions
@@ -1647,8 +1647,10 @@ class QuarkOCP_MX_MoEMethod(QuarkMoEMethod):
# If no native backend available, use emulation.
if self.mxfp4_backend is Mxfp4MoeBackend.NONE:
self.mxfp4_backend = Mxfp4MoeBackend.EMULATION
self.experts_cls = None
self.experts_cls = backend_to_kernel_cls(self.mxfp4_backend)[0]
if self.experts_cls is None:
self.experts_cls = backend_to_kernel_cls(self.mxfp4_backend)[0]
logger.info_once(
f"Using {self.mxfp4_backend.value} backend for {self.ocp_mx_scheme}"