import acl 

import torch 

import torch_npu 

import time 

 

acl.init() 

acl.rt.set_device(0) 

acl.mdl.load_from_file("./ocr_det2.om") 

device = torch.device('npu') 

x = torch.randn(1024,1024,dtype=torch.float16).to('npu') 

start_time_npu = time.time() 

result_npu = torch.mm(x, x) 

print(x.device) 

end_time_npu = time.time() 

print(f"NPU时间: {end_time_npu - start_time_npu}秒") 

acl初始化后，将张量转到npu上就会报错：

RuntimeError: Initialize:torch_npu/csrc/core/npu/sys_ctrl/npu_sys_ctrl.cpp:139 NPU error, error code is 100002 

[Error]: Repeated initialization or repeated loading. 

        Check whether the corresponding interface is repeatedly invoked for initialization or loading. 

报错初步看是重复初始化引起的，但我在acl初始化后，该如何将张量转到npu上呢？