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
[Python] GPU not working python package #1028
Comments
@up201007037 how did you install the python package ? |
@up201007037 please make sure that the gpu support was enabled via |
@guolinke I installed via github. Then I did OSError: /home/carlos/anaconda2/lib/python2.7/site-packages/lightgbm/lib_lightgbm.so: symbol clCreateCommandQueueWithProperties, version OPENCL_2.0 not defined in file libOpenCL.so.1 with link time reference Thank you so much, and I am sorry for the dumb question. |
checking your openGL version using make sure you are using the same version of header file and GPU drivers. and try to compile from source: |
I ran I do not have the directory lib64 under /usr. I do have usr/lib32/nvidia-384 but your command does not return results under that folder. I built from source. I did: git clone --recursive https://github.com/Microsoft/LightGBM Then I went to the python folder and ran: python setup.py install --precompile I can import and run lightgbm in python, but not with the gpu it gives that error. I also tried to modify the setup.py to
And then python setup.py install --gpu I remember that I already had CUDA installed but I got some kind of error in the compile and I did sudo apt-get install libboost-all-dev and was able to compile and install. Thank you so much for your help. |
I uninstalled lightgbm via pip uninstall and then:
When i run -- The C compiler identification is GNU 5.4.0 I was able to install with sucess, but when I run:
The same error occurs: I am running that with all the defaults... This is the terminal message: `[LightGBM] [Info] This is the GPU trainer!! With that installation method I am able to run: [LightGBM] [Info] Finished loading parameters Second method: -- The C compiler identification is GNU 5.4.0 Since we do not specify the CL version we are not using the cuda version. Then: python setup.py install --gpu OSError: /home/carlos/anaconda2/lib/python2.7/site-packages/lightgbm/lib_lightgbm.so: symbol clCreateCommandQueueWithProperties, version OPENCL_2.0 not defined in file libOpenCL.so.1 with link time reference |
@up201007037 It seems that this answer #715 (comment) helped to 4 people, maybe you'll be the happy 5th one 😄 . |
@StrikerRUS I already followed all the steps described on those threads. It still does not work. This is not a problem of OpenCL I guess... I am using the features from a pandas dataframe as categorical and I am not doing one hot encoding as suggested. |
Then maybe @huanzhang12 have some thoughts about this situation. |
Thank you @StrikerRUS for your thoughts on this. |
@up201007037 the previous error that you encounter is because of that linking error, which has already solved by linking to the right opengl. |
Thank you @chivee. I would like to just give a suggestion to improve the documentation that is: if someone has CUDA installed then should run this: Thank you for your attention and help! |
Please search your question on previous issues, stackoverflow or other search engines before you open a new one.
For bugs and unexpected issues, please provide following information, so that we could reproduce on our system.
Environment info
Operating System: Ubuntu 16.04
CPU: i7, Nvidia 1060
C++/Python/R version: Python 2.7
Error Message:
LightGBMError: bin size 5858 cannot run on GPU
Reproducible examples
lgb.train({'device':'gpu'}, ds)
Steps to reproduce
Hi, I am sorry if this is already been answered but I did not find the answer for this. I was able to install the gpu version of lightgbm and I ran this with sucess: ./lightgbm config=lightgbm_gpu.conf data=higgs.train valid=higgs.test objective=binary metric=auc
However, with all the defaults, when I try to use the python package, with gpu support, this error occurs. Any idea of what might be causing this? Thank you.
The text was updated successfully, but these errors were encountered: