Skip to content

Add tip for adding nvcc to a working torch w/ conda #206

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.

Sign up for GitHub

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

Open
HastingsGreer wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Add tip for adding nvcc to a working torch w/ conda #206

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
19aa675
Add tip for adding nvcc to a working torch w/ conda
HastingsGreer Dec 19, 2021
972c7c8
Update installation.rst
HastingsGreer Dec 21, 2021
2563669
Update installation.rst
HastingsGreer Dec 21, 2021
62b390b
fix bash code block
HastingsGreer Dec 21, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 23 additions & 0 deletions doc/python/installation.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,6 +149,29 @@ First of all, make sure that you are using a C++ compiler which is compatible wi

.. _`part.cache`:

Adding nvcc to a working installation of pytorch
-------------------------------------------------
A common starting point for users looking to get started with pykeops, is to have a working gpu accelerated installation of pytorch, without nvcc. In this situation, it is possible to add nvcc to PATH without installing anything globally or modifying your drivers.

1. Find the version of CUDA that pytorch is using. With conda, this is often different from the system-wide CUDA version

.. code-block:: python

import torch
torch.version.cuda

2. Download a matching CUDA runfile installer from https://developer.nvidia.com/cuda-toolkit-archive

3. Install the runfile using the --toolkit and --toolkitpath options. This installs nvcc to a directory, without modifying your drivers or system configuration. Do not use sudo, to guarantee that this step doesn't modify your drivers. For example,

.. code-block:: bash

sh cuda_version_here_linux --silent --override --toolkit --toolkitpath=/local/path/to/put/nvcc

4. Add /local/path/to/put/nvcc/bin to $PATH and /local/path/to/put/nvcc/lib64 to $LD_LIBRARY_PATH . You can either do this in .bashrc, or a script local to your pykeops project



Cache directory
---------------

Expand Down