search

DeepGraphLearning / graphvite

GraphVite: A General and High-performance Graph Embedding System
https://graphvite.io
Apache License 2.0
1.22k stars 151 forks source link

CUDA version error #15

Closed joshuia closed 4 years ago

joshuia commented 5 years ago

Hi, my env is: conda list graphvite Name Version Build Channel graphvite 0.1.0 py27cuda92 milagraph nvcc -V nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2018 NVIDIA Corporation Built on Wed_Apr_11_23:16:29_CDT_2018 Cuda compilation tools, release 9.2, V9.2.88

But when I run graphvite baseline quick start it shows: running baseline: quick_start.yaml loading graph from /home/users/shiyixuan/.graphvite/dataset/blogcatalog/blogcatalog_train.txt 0.00018387% <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< Graph ------------------ Graph -------------------

vertex: 10312, #edge: 333983

as undirected: yes, normalization: no

[time] GraphApplication.load: 0.512994 s Check failed: error == cudaSuccess CUDA error CUDA driver version is insufficient for CUDA runtime version at /lustre04/scratch/zhuzhaoc/conda/conda-bld/graphvite_1565206157156/work/include/core/solver.h:182 Check failure stack trace: @ 0x7f740a1144dd google::LogMessage::Fail() @ 0x7f740a11c071 google::LogMessage::SendToLog() @ 0x7f740a113ecd google::LogMessage::Flush() @ 0x7f740a11576a google::LogMessageFatal::~LogMessageFatal() @ 0x7f7408d290b9 graphvite::CudaCheck() @ 0x7f7408e41105 graphvite::SolverMixin<>::SolverMixin() @ 0x7f7408e7f6b5 _ZZN8pybind1112cpp_function10initializeIZNS_6detail8initimpl11constructorIJSt6vectorIiSaIiEEimEE7executeINS_6class_IN9graphvite11GraphSolverILm128EfjEEJEEEJNS_10call_guardIJNS_18gil_scoped_releaseEEEENS_5arg_vESI_SI_ELi0EEEvRT_DpRKT0_EUlRNS2_16value_and_holderES7_imE_vJSQ_S7_imEJNS_4nameENS_9is_methodENS_7siblingENS2_24is_new_style_constructorESH_SI_SI_SI_EEEvOSJ_PFT0_DpT1_EDpRKT2_ENUlRNS2_13function_callEE1_4FUNES17 @ 0x7f7408dae469 pybind11::cpp_function::dispatcher() @ 0x7f7411f8d973 PyObject_Call @ 0x7f7411f9c80d instancemethod_call @ 0x7f7411f8d973 PyObject_Call @ 0x7f7411fe6764 slot_tp_init

Is that means cuda version error? And "/lustre04/scratch/zhuzhaoc/conda/conda-bld/graphvite_1565206157156/work/include/core/solver.h", this path is not mine.

KiddoZhu commented 5 years ago

It seems that conda always install the latest cudatoolkit by default. Try the following line

conda install -c milagraph graphvite cudatoolkit=9.2
KiddoZhu commented 5 years ago

Duplicate of #11

joshuia commented 5 years ago

Thanks, But it seems not cudatoolkit version? conda list cudatoolkit shows: cudatoolkit 9.2 0 defaults

joshuia commented 5 years ago

11 is my issue too, sorry for that I closed it incorrectly

KiddoZhu commented 5 years ago

Could you show me your version of cudatoolkit and graphvite?

To my knowledge, they should be the same CUDA version according to conda's mechanism. Specifying the version for cudatoolkit will result in the corresponding version of graphvite.

BuddhaZhang commented 5 years ago

Hello, I met same question.

conda list graphvite
# packages in environment at /home/zrg/anaconda3:
#
# Name           Version               Build Channel
graphvite       0.1.0               py37cuda92hbecf2fe    milagraph
conda list cudatoolkit
# packages in environment at /home/zrg/anaconda3:
#
# Name           Version               Build Channel
cudatoolkit      9.2                       0
zjduan commented 5 years ago

@joshuia What's your CUDA Driver Version? You can check it by command cat /proc/driver/nvidia/version. Each release of the CUDA Toolkit requires a minimum version of the CUDA driver. Maybe this is the reason.