[Question/Issue] How to install cuquantum on WSL2 Ubuntu-20.04? #13
Answered
by
Sinestro38
Sinestro38
asked this question in
Q&A
-
|
I'm unsure whether this is an issue on my side, or just a missing compatibility. with the following cuBLAS error: Any idea what may be causing this? I looked into cublas directly but I thought it may be too invasive for such a simple issue. I already have custatevec working on an ubuntu GCP instance so I'm wondering if this is due to an incompatibility with Window's Subsystem for Linux 2. |
Beta Was this translation helpful? Give feedback.
Answered by
Sinestro38
Aug 15, 2022
Replies: 10 comments 1 reply
-
|
Hi @Sinestro38, thanks for reporting. While WSL2 is not officially supported by cuQuantum, we have several successful reports, so let me try to understand the problem better. Do you already have CUDA installed in your WSL2 environment? In particular, have you followed the Linux-Ubuntu instruction in https://developer.nvidia.com/cuda-downloads to set up the network repo? This error is likely telling us that CUDA is not installed via apt, and the repo is not set up, so apt cannot resolve the package dependencies for us. |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Correction: CUDA-WSL2 instruction is here: https://docs.nvidia.com/cuda/wsl-user-guide/index.html#cuda-support-for-wsl2 |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
@leofang Yes, I had installed the CUDA toolkit following their specific instructions for WSL2-Ubuntu a few months prior when I got the laptop. This is the output from running |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
@Sinestro38 could you attach the outputs of the following two commands?
|
Beta Was this translation helpful? Give feedback.
0 replies
-
If you check above, it does say that |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Hi @Sinestro38, @JeremyWangNVDA noticed that you're using the local deb installation, not the network deb installation, could you give it a shot? For reference I copy/paste the instruction here: |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Ah, it almost worked! No idea why the local installation didn't work though. This time it recognized the I tried running a custatevec code sample and it didn't work either (didn't recognize Any idea where I need to put the symlinks, and from where to where? Thanks! @leofang |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
It seems this error is a known WSL issue. I'd kindly direct you to the following links for potential workarounds:
with a heads-up that this is use-at-own-risk. I will pass this around internally and see if the team has a better solution. Personally I'd suggest WSL2 users to either use |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Update: tarballs are available here: https://developer.download.nvidia.com/compute/cuquantum/redist/cuquantum/linux-x86_64/ |
Beta Was this translation helpful? Give feedback.
-
|
Let me change this issue to a GH discussion. |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Thanks @leofang , I got the network installation to work!
Thought I'll post the full set of steps I followed here in case anyone finds it of use. Crucially, I added the last 2 of the following 4 lines to the top of my And then the code sample worked right afterwards! |
Beta Was this translation helpful? Give feedback.
0 replies
Answer selected by
leofang
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Thanks @leofang , I got the network installation to work!
Thought I'll post the full set of steps I followed here in case anyone finds it of use.
Crucially, I added the last 2 of the following 4 lines to the top of my
.bashrcso that the cuquantum files are included in theLD_LIBRARY_PATH.