detectron2 运行正常，但 tensormask 一直显示 RuntimeError： Not compiled with GPU support

问：

对不起，我的英语不好。 我的操作系统是 WIN11，我使用 Anaconda 创建了一个虚拟环境，python 版本是 3.8.18，并使用

conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge

安装火炬和 CUDA。

然后cd到detectron2文件夹，pip安装并运行一个 demo.py，程序运行非常流畅。 虽然“print（torch.cuda.is_available（）， CUDA_HOME）”的显示显示（True，None），但它仍然可以运行我的训练代码。

之后，我将 cd 到 tensormask 文件夹并 pip 安装它，它显示“已成功安装 tensormask-0.1”，但是当我尝试运行我的训练代码时，它一直显示

Y = \_C.swap_align2nat_forward(X, lambda_val, pad_val)
RuntimeError: Not compiled with GPU support

.

整个错误是：

ERROR \[11/17 06:01:47 d2.engine.train_loop\]: Exception during training:
Traceback (most recent call last):
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\detectron2\\engine\\train_loop.py", line 155, in train
self.run_step()
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\detectron2\\engine\\defaults.py", line 496, in run_step
self.\_trainer.run_step()
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\detectron2\\engine\\train_loop.py", line 310, in run_step    loss_dict = self.model(data)
File "C:\\Users\\andy5\\anaconda3\\envs\\detect\\lib\\site-packages\\torch\\nn\\modules\\module.py", line 1102, in \_call_impl
return forward_call(\*input, \*\*kwargs)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\projects\\tensormask\\tensormask\\arch.py", line 382, in forward
pred_logits, pred_deltas, pred_masks = self.head(features)
File "C:\\Users\\andy5\\anaconda3\\envs\\detect\\lib\\site-packages\\torch\\nn\\modules\\module.py", line 1102, in \_call_impl
return forward_call(\*input, \*\*kwargs)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\projects\\tensormask\\tensormask\\arch.py", line 908, in forward
cur_mask = cur_mask_module(cur_mask)
File "C:\\Users\\andy5\\anaconda3\\envs\\detect\\lib\\site-packages\\torch\\nn\\modules\\module.py", line 1102, in \_call_impl
return forward_call(\*input, \*\*kwargs)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\projects\\tensormask\\tensormask\\layers\\swap_align2nat.py", line 54, in forward
return swap_align2nat(X, self.lambda_val, self.pad_val)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\projects\\tensormask\\tensormask\\layers\\swap_align2nat.py", line 15, in forward
Y = \_C.swap_align2nat_forward(X, lambda_val, pad_val)
RuntimeError: Not compiled with GPU support
\[11/17 06:01:47 d2.engine.hooks\]: Total training time: 0:00:03 (0:00:00 on hooks)
\[11/17 06:01:47 d2.utils.events\]:  iter: 0       lr: N/A  max_mem: 3872M
Traceback (most recent call last):
File "Tensormask_Train.py", line 37, in \<module\>
trainer.train()
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\detectron2\\engine\\defaults.py", line 486, in train
super().train(self.start_iter, self.max_iter)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\detectron2\\engine\\train_loop.py", line 155, in train
self.run_step()
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\detectron2\\engine\\defaults.py", line 496, in run_step
self.\_trainer.run_step()
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\detectron2\\engine\\train_loop.py", line 310, in run_step    loss_dict = self.model(data)
File "C:\\Users\\andy5\\anaconda3\\envs\\detect\\lib\\site-packages\\torch\\nn\\modules\\module.py", line 1102, in \_call_impl
return forward_call(\*input, \*\*kwargs)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\projects\\tensormask\\tensormask\\arch.py", line 382, in forward
pred_logits, pred_deltas, pred_masks = self.head(features)
File "C:\\Users\\andy5\\anaconda3\\envs\\detect\\lib\\site-packages\\torch\\nn\\modules\\module.py", line 1102, in \_call_impl
return forward_call(\*input, \*\*kwargs)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\projects\\tensormask\\tensormask\\arch.py", line 908, in forward
cur_mask = cur_mask_module(cur_mask)
File "C:\\Users\\andy5\\anaconda3\\envs\\detect\\lib\\site-packages\\torch\\nn\\modules\\module.py", line 1102, in \_call_impl
return forward_call(\*input, \*\*kwargs)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\projects\\tensormask\\tensormask\\layers\\swap_align2nat.py", line 54, in forward
return swap_align2nat(X, self.lambda_val, self.pad_val)
File "c:\\users\\andy5\\downloads\\detectron2-main\\detectron2-main\\projects\\tensormask\\tensormask\\layers\\swap_align2nat.py", line 15, in forward
Y = \_C.swap_align2nat_forward(X, lambda_val, pad_val)
RuntimeError: Not compiled with GPU support

我搜索了很多问题，但大多数都与 Detectron2 训练期间的“RuntimeError： Not compiled with GPU support error”有关。此外，官方建议是用print（torch.cuda.is_available（），CUDA_HOME）确认。

但是，我面临的问题是 Detectron2 训练和下载 TensorMask（with （torch.cuda.is_available， CUDA_HOME） show （True， None）） 一切正常，但只有在我尝试使用 TensorMask 进行训练时才会发生错误。

我该如何解决这个问题，或者是否有任何关键字可以供我搜索？