WIN10 RuntimeError: Not compiled with GPU support

[img5d]

❓ Questions and Help

Hey , guys, I`m a beginner.I need a help.

When i tried maskrcnn-benchmark, at last i encountered the issues as the title: RuntimeError: Not compiled with GPU support (nms at c:\users\gs63vr\train\maskrcnn-benchmark\maskrcnn_benchmark\csrc\nms.h:22) (no backtrace available) [ WARN:0] terminating async callback

The method i used is Step-by-step installation (Anaconda3→Creating Virtual Environment →ipython→ maskrcnn_benchmark and coco api dependencies → conda install pytorch torchvision cudatoolkit=9.0 -c pytorch → install pycocotools → apex → download maskrcnn-benchmark → revise SigmodiFocalLoss_cuda.cu → instal maskrcnn-benchmark , run demo test )

and my enviroment is win10. Pytorch 1.1 py3.6 cuda90 cudnn7.1 torchvision 0.2.2

The whole information as below:

Traceback (most recent call last): File "webcam.py", line 80, in main() File "webcam.py", line 71, in main composite = coco_demo.run_on_opencv_image(img) File "C:\Users\GS63VR\train\maskrcnn-benchmark\demo\predictor.py", line 172, in run_on_opencv_image predictions = self.compute_prediction(image) File "C:\Users\GS63VR\train\maskrcnn-benchmark\demo\predictor.py", line 205, in compute_prediction predictions = self.model(image_list) File "C:\Users\GS63VR\Anaconda3\envs\test\lib\site-packages\torch\nn\modules\module.py", line 493, in call result = self.forward(*input, kwargs) File "c:\users\gs63vr\train\maskrcnn-benchmark\maskrcnn_benchmark\modeling\detector\generalized_rcnn.py", line 50, in forward proposals, proposal_losses = self.rpn(images, features, targets) File "C:\Users\GS63VR\Anaconda3\envs\test\lib\site-packages\torch\nn\modules\module.py", line 493, in call result = self.forward(*input, *kwargs) File "c:\users\gs63vr\train\maskrcnn-benchmark\maskrcnn_benchmark\modeling\rpn\rpn.py", line 161, in forward return self._forward_test(anchors, objectness, rpn_box_regression) File "c:\users\gs63vr\train\maskrcnn-benchmark\maskrcnn_benchmark\modeling\rpn\rpn.py", line 187, in _forward_test boxes = self.box_selector_test(anchors, objectness, rpn_box_regression) File "C:\Users\GS63VR\Anaconda3\envs\test\lib\site-packages\torch\nn\modules\module.py", line 493, in call result = self.forward(input, kwargs) File "c:\users\gs63vr\train\maskrcnn-benchmark\maskrcnn_benchmark\modeling\rpn\inference.py", line 140, in forward sampled_boxes.append(self.forward_for_single_feature_map(a, o, b)) File "c:\users\gs63vr\train\maskrcnn-benchmark\maskrcnn_benchmark\modeling\rpn\inference.py", line 120, in forward_for_single_feature_map score_field="objectness", File "c:\users\gs63vr\train\maskrcnn-benchmark\maskrcnn_benchmark\structures\boxlist_ops.py", line 27, in boxlist_nms keep = _box_nms(boxes, score, nms_thresh) File "C:\Users\GS63VR\Anaconda3\envs\test\lib\site-packages\apex-0.1-py3.6.egg\apex\amp\amp.py", line 22, in wrapper RuntimeError: Not compiled with GPU support (nms at c:\users\gs63vr\train\maskrcnn-benchmark\maskrcnn_benchmark\csrc\nms.h:22) (no backtrace available) [ WARN:0] terminating async callback

[jefflomax]

I haven't run webcam.py in Windows 10 yet, but my setup/steps to run train.py are below. It was a bit of a struggle, if I've left things out I'm sure we can figure it out:

MaskRCNNWindowsSetup.txt

[img5d]

I haven't run webcam.py in Windows 10 yet, but my setup/steps to run train.py are below. It was a bit of a struggle, if I've left things out I'm sure we can figure it out: MaskRCNNWindowsSetup.txt

I have a question before install pytorch it should be pre-install the CUDA? If it is , may be this lead the question.

[jefflomax]

I installed CUDA first, and Visual Studio 2017 with the CUDA support. I would suggest you decide what CUDA version you'll want (9 or 10) from the choices in "Quick Start Locally" https://pytorch.org/

I just used Preview (Nightly) and PIP on Windows 10. Pick what you like best.

On Sun, May 12, 2019 at 11:19 PM img5d notifications@github.com wrote:

I haven't run webcam.py in Windows 10 yet, but my setup/steps to run train.py are below. It was a bit of a struggle, if I've left things out I'm sure we can figure it out: MaskRCNNWindowsSetup.txt

I have a question before install pytorch it should be pre-install the CUDA? If it is , may be this lead the question.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/facebookresearch/maskrcnn-benchmark/issues/757#issuecomment-491670695, or mute the thread https://github.com/notifications/unsubscribe-auth/ABLJKZ6M4GZ5XYRHUEP67DLPVDT5FANCNFSM4HLNN7NQ .