RuntimeError: THPDevice_Check(map_location.ptr())INTERNAL ASSERT FAILED

Question

一、问题现象：

Traceback (most recent call last): 
  File "/root/anaconda3/envs/Hpytorch/models/IEEE_TGRS_LDGnet-main/train_queue.py", line 216, in <module> 
    pretrained_dict  = torch.jit.load('./ViT-B-32.pt', map_location="cpu").state_dict() 
  File "/root/anaconda3/envs/Hpytorch/lib/python3.9/site-packages/torch/jit/_serialization.py", line 162, in load 
    cpp_module = torch._C.import_ir_module(cu, str(f), map_location, _extra_files) 
RuntimeError: THPDevice_Check(map_location.ptr())INTERNAL ASSERT FAILED at "/root/pytorch/torch/csrc/jit/python/script_init.cpp":1769, please report a bug to PyTorch. 

二、软件版本:
-- CANN 版本: 8.0RC1
-- Pytorch 版本: 1.11.0
-- Python 版本: 3.9.16
-- 操作系统版本: Ubuntu 22.04

三、测试步骤：
1.在昇腾服务器中进行模型自动迁移
2.在训练脚本中输入import torch_npu和from torch_npu.contrib import transfer_to_npu
3.运行训练脚本
4.出现NotlmplementedError: Could not run 'aten::empty strided' with arguments from the 'CUDA" backend
5.修改训练脚本中的pretrained dict = torch.jit.load('./viT-B-32.pt', map location="cpu").state dict()，加了jit
6.然后报了如一bug

想问问是怎么回事

Answer 1

以下回复参考：皆我百晓生、券券喵儿等免费微信小程序作答:

根据给定的错误信息，可以看出