Torch not compiled with CUDA enabled

问题遇到的现象和发生背景

在本地不含GPU的电脑上跑torch，想把包含cuda的部分去掉。

问题相关代码，请勿粘贴截图

""" Example for biomarker identification
"""
import os
import copy
from feat_importance import cal_feat_imp, summarize_imp_feat
from utils import load_model_dict

if __name__ == "__main__":
    data_folder = 'BRCA'
    model_folder = os.path.join(data_folder, 'models')
    view_list = [1,2,3]
    if data_folder == 'ROSMAP':
        num_class = 2
    if data_folder == 'BRCA':
        num_class = 5

    featimp_list_list = []
    for rep in range(5):
        featimp_list = cal_feat_imp(data_folder, os.path.join(model_folder, str(rep+1)), 
                                    view_list, num_class)
        featimp_list_list.append(copy.deepcopy(featimp_list))
    summarize_imp_feat(featimp_list_list)
    

运行结果及报错内容

AssertionError                            Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_11144/1698963967.py in <module>
     18     for rep in range(5):
     19         featimp_list = cal_feat_imp(data_folder, os.path.join(model_folder, str(rep+1)), 
---> 20                                     view_list, num_class)
     21         featimp_list_list.append(copy.deepcopy(featimp_list))
     22     summarize_imp_feat(featimp_list_list)

D:\python\Book\MOGONET-main\feat_importance.py in cal_feat_imp(data_folder, model_folder, view_list, num_class)
     33         if cuda:
     34             model_dict[m].cuda()
---> 35     model_dict = load_model_dict(model_folder, model_dict)
     36     te_prob = test_epoch(data_trte_list, adj_te_list, trte_idx["te"], model_dict)
     37     if num_class == 2:

D:\python\Book\MOGONET-main\utils.py in load_model_dict(folder, model_dict)
    127         if os.path.exists(os.path.join(folder, module+".pth")):
    128 #            print("Module {:} loaded!".format(module))
--> 129             model_dict[module].load_state_dict(torch.load(os.path.join(folder, module+".pth"), map_location=torch.device('cpu')))
    130         else:
    131             print("WARNING: Module {:} from model_dict is not loaded!".format(module))

D:\Ancondada\envs\torch\lib\site-packages\torch\cuda\__init__.py in current_device()
    477 def current_device() -> int:
    478     r"""Returns the index of a currently selected device."""
--> 479     _lazy_init()
    480     return torch._C._cuda_getDevice()
    481 

D:\Ancondada\envs\torch\lib\site-packages\torch\cuda\__init__.py in _lazy_init()
    206                 "multiprocessing, you must use the 'spawn' start method")
    207         if not hasattr(torch._C, '_cuda_getDeviceCount'):
--> 208             raise AssertionError("Torch not compiled with CUDA enabled")
    209         if _cudart is None:
    210             raise AssertionError(

AssertionError: Torch not compiled with CUDA enabled

我的解答思路和尝试过的方法

把utils中的 load_model_dict（）函数中的map_location改成使用 cpu（上面的已经改过），还是不行

我想要达到的结果

程序能够成功运行

首先确保你装的torch是cpu版本