python build_pkg.py failed #19
Comments
|
I am facing the same problem, I tried with Ubuntu, Python and Pytorch/Pytorch base images for Docker and they all seem to have the same issue. Were you able to solve it ? |
|
Thank you very much, that worked. But when I try to train or run `[3/3] c++ emd.o emd_kernel.cuda.o -shared -L/opt/conda/lib/python3.8/site-packages/torch/lib -lc10 -lc10_cuda -ltorch_cpu -ltorch_cuda_cu -ltorch_cuda_cpp -ltorch -ltorch_python -L/usr/local/cuda/lib64 -lcudart -o emd_ext.so Loading extension module emd_ext... load emd_ext time: 47.402s` How long should I wait after this part ? does it usually take long? thank you in advance! |
|
From the log it seems it finished the building already. The building time varies (on my side it takes ~0.1s to build). If you dont need to compute the evaluation metric and want to skip build EMD, one dirty solution is not using EMD at all and turn it off by 1) comment out this line and this line and 2) overwrite this function to make it return any constant. |
@ZENGXH I'm running into this exact issue that my build/training gets stuck after it has loaded emd. It says it loaded emd_ext after around ~0.3 seconds. Afterwards it continues to hang without any response. |
|
@JohanYe if you control+c to kill the program, can you see where the program is hanging? For example, from the killed message in the code? |
|
For the |
This is the message from the keyboard interrupt. Now i am unable to get it to run at all. This occurs both during build_pkg.py and training. So it prevents me running any code @ZENGXH |
|
With the hints from the error i was able to track down the issue. It seems to be solved by deleting the folder: "LION/third_party/ChamferDistancePytorch/tmp". Similarly to this issue: zhou13/neurvps#1 (comment) Note that the github issue contains help to find the exact path to the cache folder that needs be deleted in case your path may differ. |
|
The same problems. |
|
Same problem, both in conda environment and in docker container. `# python build_pkg.py The above exception was the direct cause of the following exception: Traceback (most recent call last): |
|
sorry for the building error you meet. I also dont have very clear idea what's going on :( could you try to see if you can
if the pvcnn repo is run-able, it's likely the |
|
This problem is typically caused by a mismatch of your cuda version and the torch version. You can check them by and I think that CUDA 11.6 is required by this repo's code. |
|
the problem is still exist, and can i use the demo.py normally without any gpu framework? |
|
I didn't try cpu only before. May be hard to get it work without gpu |
|
Can i execute this code on Windows? |
I didn't test it on Windows. If you could install the package it's possible |
Hi, I've been trying to run the set-up steps to do some training locally, but I'm stuck on the last optional step, and even with this the demo or training doesn't run. I ran the set up steps on WSL Ubuntu, but when I run python
build_pkg.pyit fails after the the line below. I also tried runningpython demo.pywhich also failed. Woud you happen you have a docker image for inference, it would be super helpful! I tried installing CUDA 11.6 locally as well to no avail. Any help would be much appreciated!The text was updated successfully, but these errors were encountered: