Not compiled with GPU support (nms at /home/yuezhangw/Program/Faster_RCNN_torch101/faster-rcnn.pytorch/lib/model/csrc/nms.h:22)

[yuezhangW]

The processing of compiling is successful, without errors but a warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++

Then, when I try train_val.py, errors occured: Not compiled with GPU support (nms at /home/yuezhangw/Program/Faster_RCNN_torch101/faster-rcnn.pytorch/lib/model/csrc/nms.h:22)

it seems like that i had not compiled nms into GPU env, but the compiling was no error. i used py36 to complie, maybe i should try py27?As a new hand, I don't know how to use a different version Python in one env, one for complie, one for run program?

SO, anyone had the same problem?Thanks for the helping!

[loolzaaa]

Do you have _C........pyd file after compiling in /lib or more deep folder?

[yuezhangW]

Do you have _C........pyd file after compiling in /lib or more deep folder?

yes, but it didn't work

[loolzaaa]

It seems that code can't detect CUDA device. torch.cuda.is_available() return True for you?

If yes, try to remove CUDA_HOME from this line.

Of course, i hope you use PyTorch 1.0 branch.

[yuezhangW]

It seems that code can't detect CUDA device. torch.cuda.is_available() return True for you?

If yes, try to remove CUDA_HOME from this line.

Of course, i hope you use PyTorch 1.0 branch.

maybe you are right, i will try it tomorrow! Much thanks to your reply!

[yuezhangW]

It seems that code can't detect CUDA device. torch.cuda.is_available() return True for you?

If yes, try to remove CUDA_HOME from this line.

Of course, i hope you use PyTorch 1.0 branch.

yes, I used Pytorch1.0 branch,the readme file can prove it if torch.cuda.is_available() and CUDA_HOME is not None I try to remove part of this, and it turns to

if torch.cuda.is_available() , torch.cuda.is_available() always returns Ture

then I run the file setup.py again, error occurs:

Traceback (most recent call last): File "setup.py", line 65, in ext_modules=get_extensions(), File "setup.py", line 52, in get_extensions extra_compile_args=extra_compile_args, File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 476, in CUDAExtension library_dirs += library_paths(cuda=True) File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 549, in library_paths paths.append(_join_cuda_home(lib_dir)) File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 1121, in _join_cuda_home raise EnvironmentError('CUDA_HOME environment variable is not set. ' OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root. (py36_torch101_cuda90) yuezhangw@yuezhangw-PC:~/Program/Faster_RCNN_torch101/faster-rcnn.pytorch/lib$ python setup.py build develop Traceback (most recent call last): File "setup.py", line 65, in ext_modules=get_extensions(), File "setup.py", line 52, in get_extensions extra_compile_args=extra_compile_args, File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 476, in CUDAExtension library_dirs += library_paths(cuda=True) File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 549, in library_paths paths.append(_join_cuda_home(lib_dir)) File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 1121, in _join_cuda_home raise EnvironmentError('CUDA_HOME environment variable is not set. ' (py36_torch101_cuda90) yuezhangw@yuezhangw-PC:/Program/Faster_RCNN_torch101/faster-rcnn.pytorch/lib$ python setup.py build develop yes Traceback (most recent call last): File "setup.py", line 66, in ext_modules=get_extensions(), File "setup.py", line 53, in get_extensions extra_compile_args=extra_compile_args, File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 476, in CUDAExtension library_dirs += library_paths(cuda=True) File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 549, in library_paths paths.append(_join_cuda_home(lib_dir)) File "/home/yuezhangw/anaconda3/envs/py36_torch101_cuda90/lib/python3.6/site-packages/torch/utils/cpp_extension.py", line 1121, in _join_cuda_home raise EnvironmentError('CUDA_HOME environment variable is not set. ' OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root.

do you know how to deal with it?Thank you ever so much

[loolzaaa]

Ok, i see. You use anaconda and i suppose you install CUDA through it, so it is not the entire package.

Install CUDA drivers manually https://developer.nvidia.com/cuda-downloads

After installation, pytorch would be able to access the CUDA path. You can check it:

import os
print(os.environ.get('CUDA_PATH')

[Chenlei6]

Excuse me! I meet the same problem with yours, do you solve it now? if yes, please share the solution, thanks!

[yuezhangW]

Excuse me! I meet the same problem with yours, do you solve it now? if yes, please share the solution, thanks!

sorry,i did not solve it till now

[xxxjw-fighting]

Before compiling setup.py, you should make sure that print(torch.cuda.is_available()) returns True. I've solved this problem!