Closed stromal closed 4 years ago
It is an incorrect answer I have tried that as well and it does not works.
conda install -c nvidia -c rapidsai py-xgboost
I left my answer in the forum post.
https://discuss.xgboost.ai/t/xgboost-version-not-compiled-with-gpu-support/1936/4?u=hcho3
I was struggling to run XGBoost on AWS sagemaker GPU.
But conda install py-xgboost-gpu
worked as as charm for me!
Thanks to @bonyii, on this issue here.
Still having same issue, even after following @GoulartNogueira. Do you need to uninstall any of the below after installing py-xgboost-gpu? My current installed libraries are: py-xgboost-mutex 2.0 gpu_0 conda-forge libxgboost 1.7.6 cuda120ha80de60_5 conda-forge py-xgboost 1.7.6 cuda120py311h647eaca_5 conda-forge py-xgboost-gpu 1.7.6 py311h0f3d150_5 conda-forge xgboost 1.7.6 cuda120py311h647eaca_5 conda-forge
check if you still have conflicting libraries between pip and conda, i had a similar issue, I used my other pc that only had conda and it worked. you can also try your code on collab/Kaggle, remember to set the GPU runtime. also make sure you've installed you gpu drivers and make sure you've cuda installed.
CODE
it is a kaggle project that I am trying to run locally on GPU: GTX 1080, Ubuntu 20.04 TS
ERROR
INPUT
OUTPUT
MY LIBRARIES
I have installed the following but they did not helped: