Open chenzhanjie opened 5 years ago
Have you installed cuda successfully?
Have you installed cuda successfully?
Yes, i created an environment in Anaconda3, and then i installed the pytorch1.0 with cuda9.0 and python3.6.
What is your output of nvidia-smi
, nvcc --version
, and
python
import torch
torch.__version__
?
What is your output of
nvidia-smi
,nvcc --version
, andpython
import torch
torch.__version__
?
these output you mentioned are right, but now i think the problem may be caused by network connection, I am in China. So just like you said,it is possible that i haven't installed it completely. But I do not find which part i missed yet. So I am trying to reinstall all the components once.
i met the same error when i run python demo.py. and my nvcc -V is v10.0(titan rtx), and my torch version is 1.0.0, torchvision is 0.2.2. should the titan rtx install with cuda10.0? like conda install pytorch=1.0.0 cuda100 -c pytorch
after run conda install pytorch=1.0.0 cuda100 -c pytorch and conda install torchvision -c pytorch, it worked for me, wish it may help you : )
I run the project in pytorch 1.0 version, and then the result have been print. But there is a problem in the terminal: THCudaCheck FAIL file=/opt/conda/conda-bld/pytorch_1544174967633/work/aten/src/THC/THCGeneral.cpp line=405 error=11 : invalid argument I did not find the answer, so I hope that I can get some help at here. Thank you.