[Compile][Platform] Make PiecewiseBackend pluggable and extendable (#18076)

Signed-off-by: Mengqing Cao <cmq0113@163.com> Co-authored-by: youkaichao <youkaichao@gmail.com>
2025-05-23 03:11:53 +08:00
parent 721fb9b181
commit f8d2cc5f55
6 changed files with 305 additions and 200 deletions
--- a/vllm/platforms/rocm.py
+++ b/vllm/platforms/rocm.py
@@ -382,3 +382,7 @@ class RocmPlatform(Platform):
    @classmethod
    def is_navi(cls) -> bool:
        return 'gfx1' in torch.cuda.get_device_properties(0).gcnArchName
+
+    @classmethod
+    def get_piecewise_backend_cls(cls) -> str:
+        return "vllm.compilation.cuda_piecewise_backend.CUDAPiecewiseBackend"  # noqa