[XPU][NPU][ROCM] enable cpu_offloading flag for non_cuda (#1488)

Signed-off-by: gcanlin <canlinguosdu@gmail.com>
Signed-off-by: Chendi Xue <chendi.xue@intel.com>
Co-authored-by: gcanlin <canlinguosdu@gmail.com>

vllm_omni/diffusion/offloader/sequential_backend.py

@@ -76,9 +76,11 @@ class SequentialOffloadHook(ModelHook):
         current_omni_platform.synchronize()
 
         logger.debug(
-            "Swapped: %s -> CPU, %s -> GPU",
+            "Swapped: %s -> CPU, %s -> %s, free memory: %.4f GB",
             [t.__class__.__name__ for t in self.offload_targets],
             module.__class__.__name__,
+            f"{self.device.type}:{self.device.index}",
+            current_omni_platform.get_free_memory() / 1024 / 1024 / 1024,
         )
 
         return args, kwargs