-
Notifications
You must be signed in to change notification settings - Fork 1.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
compilation error C2131: expression did not evaluate to a constant #169
Comments
This is about the windows compatibility issue. We may not be able to fix it quickly because currently, we do not have windows machines. |
Thank you for the reply. I'll keep on it myself for a while and see if I can come up with a workaround. If I do I'll be sure to post the solution here. |
Great! Feel free to create a PR to solve it if you find any solution. Thanks in advance. |
sir,I have the same problem too.Have you find the way to slove it? @damonftl |
@huixiancheng I did not end up solving it, sorry about that. I actually went as far as installing Linux and dual-booting just so I could try this out (and it did work fine in Linux). I did manage to get VoteNet (https://github.com/facebookresearch/votenet) working in Windows with some small code adjustments. You may be able to get that one going. |
@ZwwWayne Did you find a way around it? |
Can you tell me the code adjustments you made? |
Of course it was five months ago now and I don't completely remember everything I had to do, sorry about that. I know it was not very much. Dependency versions have to be exactly correct or it won't work and you'll get weird and often misleading errors. For some reason batch sizes in Windows have to be set lower than in Linux or you'll get GPU memory errors (maybe something with nvidia's drivers? I don't know). I just searched through the code real quick and didn't see any big obvious edits. A quick google search should likely provide solutions to the problems you might hit. The really hard part was getting it to work with our custom data, but that would be true in Linux as well. |
Hi, the line which was causing error it got resolved by changing the the erroring line to dynamic array allocation at two places
The while compiling
|
I was able to fix these issues: #551 |
…en-mmlab#169) * add ppl install doc and how to test a model doc * remove mmdet model list in ppl.md * resolve comments * fix typo * add mmocr doc * mv configs as the final column * fix typo and align tables * fix lint
* support export hardsigmoid in torch<=1.8 * fix lint
have you solved this problem? i met the same problem... if you could help im really appreciate that , thx |
Hi, I'm trying to install and use mmdetection on my Windows 10 machine (I know, that's my first problem...) and I'm at literally the last step of the process. Everything appears to have installed and compiled correctly up to this point, but when running the "pip install -v -e ." inside the mmdetection3d directory I receive the following errors for iou3d:
In researching this particular error, it appears to be a problem with the compiler in Windows. The problem is line 108 mentioned in the above errors:
const int col_blocks = DIVUP(boxes_num, THREADS_PER_BLOCK_NMS);
and then its use on line 127 to initialize an array:
unsigned long long remv_cpu[col_blocks];
Apparently some compilers are okay with this and some, specifically in Windows apparently, are not. I do not know how to work around this problem. I have tried various things mentioned in places like StackOverflow, like setting col_blocks to be static, but the same error appears. I have also checked the issue queue for OpenPCDet, where the original code for iou3d came from, but nobody seems to have this problem. Does anyone know of a way around this?
The text was updated successfully, but these errors were encountered: