-
Notifications
You must be signed in to change notification settings - Fork 3
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
GPU not found #10
Comments
Dear @arneelof Thanks for the question. I'm happy to see that other people are interested in Miga. The Important note. Pip will try to reuse the wheels it built for the first installation (without CUDA). So you'll have to either clear pip's cache before the next installation or run pip with the |
Thanks for your quick response
You are right for some reason it does not install without GPU. I managed to
get it to install with a GPU (QUADRO P620) on one machine and it provided a
nice speedup (8 vs 90 s), but on a couple of machines with better GPUs it
did not install. All machines are running Ubuntu 20.04
I also tried to make a singularity image (attached definition), but get the
same problem.
Any ideas what could be wrong (I also attach the output from the
singularity log installation)
Yours
Arne
…-----------------------------------------
Arne Elofsson Science for Life Laboratory
Tel:+46-(0)70 695 1045 Stockholm University
http://bioinfo.se/ Box 1031,
Email: ***@***.*** 17121 Solna, Sweden
Twitter: https://twitter.com/arneelof
Scholar: http://scholar.google.se/citations?user=s3OCM3AAAAAJ
ORCID: 0000-0002-7115-9751
On Mon, Jun 7, 2021 at 7:08 PM Caio S. Souza ***@***.***> wrote:
Dear @arneelof <https://github.com/arneelof>
Thanks for the question. I'm happy to see that other people are interested
in Miga.
The CUDA_HOME variable should point to /usr for toolkits installed using
the Ubuntu repo. Running CUDA_HOME=/usr pip3 install miga should work.
Just to be sure, you may run pip with --verbose and check if the message Installing
with CUDA is written to the output.
*Important note.* Pip will try to reuse the wheels it built for the first
installation (without CUDA). So you'll have to either clear pip's cache
before the next installation or run pip with the --no-cache-dir option
(available for pip 6.0 or newer).
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#10 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AA3LGSPOSHWEMQLKJO3CV6TTRT4IPANCNFSM46H7DCQQ>
.
|
Dear @arneelof This could be a problem in the compilation of the GPU code. Please send me the output of this command: |
Thanks
Yours
Arne
…-----------------------------------------
Arne Elofsson Science for Life Laboratory
Tel:+46-(0)70 695 1045 Stockholm University
http://bioinfo.se/ Box 1031,
Email: ***@***.*** 17121 Solna, Sweden
Twitter: https://twitter.com/arneelof
Scholar: http://scholar.google.se/citations?user=s3OCM3AAAAAJ
ORCID: 0000-0002-7115-9751
On Tue, Jun 8, 2021 at 8:35 PM Caio S. Souza ***@***.***> wrote:
Dear @arneelof <https://github.com/arneelof>
This could be a problem in the compilation of the GPU code. Please send me
the output of this command: CUDA_HOME=/usr pip3 install miga
--no-cache-dir --verbose
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#10 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AA3LGSJ2Q5D53CD4C3NHLELTRZPELANCNFSM46H7DCQQ>
.
|
Sorry to bother, but for some reason I can not get miga to find my GPU.
Running ubuntu 20.04 with nvidia-cuda-toolkit 10.1.244
Perhaps it is something trivial such that I just have the wrong path in CUDA_HOME (I tried /usr/bin/ as well /usr/lib/nvidia-cuda-toolkit/ and some others without success.
Cuda looks OK to me:
$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Sun_Jul_28_19:07:16_PDT_2019
Cuda compilation tools, release 10.1, V10.1.243
Any idea are appreciated.
The text was updated successfully, but these errors were encountered: