[ROCm][CI][Docs] Add comment explaining TRITON_ATTN fallback for ROCm (#32835)

Signed-off-by: Andreas Karatzas <akaratza@amd.com>
This commit is contained in:
Andreas Karatzas
2026-01-22 00:11:09 -06:00
committed by GitHub
parent eb1629da24
commit a810299838

View File

@@ -88,6 +88,8 @@ def get_available_attention_backends() -> list[str]:
get_valid_backends = getattr(current_platform.__class__, "get_valid_backends", None)
if get_valid_backends is None:
if current_platform.is_rocm():
# ROCm uses Triton as its default attention backend since
# Flash Attention is not supported.
return ["TRITON_ATTN"]
else:
return ["FLASH_ATTN"]