Enable Pytorch Attention for gfx1150 (Comfy-Org#12973)

Author: lostdisc

comfy/model_management.py

@@ -400,7 +400,7 @@ def aotriton_supported(gpu_arch):
         if args.use_split_cross_attention == False and args.use_quad_cross_attention == False:
             if aotriton_supported(arch):  # AMD efficient attention implementation depends on aotriton.
                 if torch_version_numeric >= (2, 7):  # works on 2.6 but doesn't actually seem to improve much
-                    if any((a in arch) for a in ["gfx90a", "gfx942", "gfx950", "gfx1100", "gfx1101", "gfx1151"]):  # TODO: more arches, TODO: gfx950
+                    if any((a in arch) for a in ["gfx90a", "gfx942", "gfx950", "gfx1100", "gfx1101", "gfx1150", "gfx1151"]):  # TODO: more arches, TODO: gfx950
                         ENABLE_PYTORCH_ATTENTION = True
                 if rocm_version >= (7, 0):
                    if any((a in arch) for a in ["gfx1200", "gfx1201"]):