910b多进程出现RuntimeError: devptr INTERNAL ASSERT FAILED

[Limerence7]

环境:

910B8卡环境
docker镜像: vllm-ascend:v0.11.0rc3
torch: 2.7.1+cpu
torch_npu: 2.7.1

部分代码如下：

// memory.py
for i in range(self.num_experts):
expert = copy.deepcopy(self.template_expert)
expert = expert.to(self.devices[i % len(self.devices)])
self.current_layer.append(expert)

for i in range(self.num_experts * (self.num_layers - self.first_dense)):
expert = copy.deepcopy(self.template_expert)
self.offloaded_storages.append(expert)
for expert in self.offloaded_storages:
expert.share_memory()

// expert.py
for dev_id in range(num_workers):
p = mp.Process(target=load_function, daemon=True, args=(
dev_id, self.config, self.current_layer, self.offloaded_storages, self.load_queue, self.flag_queue))
p.daemon = True
p.start()
self.workers.append(p)

报错如下：

[WARN]operator(),build/CMakeFiles/torch_npu.dir/compiler_depend.ts:3663:Feature is not supportted and the possible cause is that driver and firmware packages do not match.
Traceback (most recent call last):
File "", line 1, in
File "/usr/local/python3.11.13/lib/python3.11/multiprocessing/spawn.py", line 122, in spawn_main
exitcode = _main(fd, parent_sentinel)
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/python3.11.13/lib/python3.11/multiprocessing/spawn.py", line 132, in _main
self = reduction.pickle.load(from_parent)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/python3.11.13/lib/python3.11/site-packages/torch_npu/multiprocessing/reductions.py", line 46, in rebuild_npu_tensor
storage = storage_cls._new_shared_npu(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/python3.11.13/lib/python3.11/site-packages/torch_npu/utils/storage.py", line 114, in _typed_storage_new_shared_npu
return torch.UntypedStorage._new_shared_npu(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/python3.11.13/lib/python3.11/site-packages/torch_npu/utils/storage.py", line 110, in _new_shared_npu
return torch_npu._C._new_shared_npu(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: devptr INTERNAL ASSERT FAILED at "build/CMakeFiles/torch_npu.dir/compiler_depend.ts":3721, please report a bug to PyTorch. entry in cache has missing shared_ptr

附加：

断开程序后，发现程序报错在p.start()处，多进程初始化出现RuntimeError