[Bugfix] Fix sleep mode level 2 (#1376)

### What this PR does / why we need it?
For sleep mode level 2, we discarded model both weights and kv_cache,
but the problems is: When we discard weights, we also discard some
tensors representing the model state which we called
`model.named_buffers()`, such as: `running_mean / running_var` in
BatchNorm、rope cos-sin cache ... when we update weights, but forgot to
update buffers as well, this will lead to some unknown issue
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?


- vLLM version: v0.10.2
- vLLM main:
5963b98b46

---------

Signed-off-by: wangli <wangli858794774@gmail.com>

tests/ut/worker/test_worker_v1.py

@@ -258,7 +258,7 @@ class TestNPUWorker(TestBase):
         # Create worker mock
         with patch.object(NPUWorker, "__init__", lambda x, **kwargs: None):
             worker = NPUWorker()
-
+            worker._sleep_saved_buffers = {}
             # Test wake_up method
             worker.wake_up(tags=["test_tag"])
 

vllm_ascend/worker/worker_v1.py

@@ -111,6 +111,9 @@ class NPUWorker(WorkerBase):
             init_cached_hf_modules()
 
         self.profiler = self._init_profiler()
+        if sleep_mode_enabled():
+            # Buffers saved before sleep
+            self._sleep_saved_buffers: dict[str, torch.Tensor] = {}
 
     def sleep(self, level: int = 1) -> None:
         if not sleep_mode_enabled():
@@ -118,6 +121,13 @@ class NPUWorker(WorkerBase):
                 "Sleep mode is not enabled. Please compile vllm-ascend with COMPILE_CUSTOM_KERNELS=1."
             )
         free_bytes_before_sleep = NPUPlatform.mem_get_info()[0]
+        # Save the buffers before level 2 sleep
+        if level == 2:
+            model = self.model_runner.model
+            self._sleep_saved_buffers = {
+                name: buffer.cpu().clone()
+                for name, buffer in model.named_buffers()
+            }
         allocator = CaMemAllocator.get_instance()
         allocator.sleep(offload_tags=("weights", ) if level == 1 else tuple())
         free_bytes_after_sleep, total = NPUPlatform.mem_get_info()
@@ -137,6 +147,14 @@ class NPUWorker(WorkerBase):
         allocator = CaMemAllocator.get_instance()
         allocator.wake_up(tags=tags)
 
+        # Restore the buffers after level 2 sleep
+        if len(self._sleep_saved_buffers):
+            model = self.model_runner.model
+            for name, buffer in model.named_buffers():
+                if name in self._sleep_saved_buffers:
+                    buffer.data.copy_(self._sleep_saved_buffers[name].data)
+            self._sleep_saved_buffers = {}
+
     def initialize_cache(self, num_gpu_blocks: int,
                          num_cpu_blocks: int) -> None:
         self.cache_config.num_gpu_blocks = num_gpu_blocks