-
Notifications
You must be signed in to change notification settings - Fork 615
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
Add a requirement for CUDA toolkit for CUDA 12 builds #4588
Conversation
!build |
CI MESSAGE: [7055526]: BUILD STARTED |
CI MESSAGE: [7055526]: BUILD PASSED |
|
||
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist --upgrade nvidia-dali-cuda110 | ||
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist --upgrade nvidia-dali-cuda120 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do you have some wording to suggest? I think the problem is, that the CUDA toolkit may be installed in several different ways depending on the system and you shouldn't mix those if you already used one of installation paths already.
Especially if you have toolkit, you don't want to repeat the install by copy-pasting it.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Generally, right now user needs to know, that he needs to install CUDA first and then DALI. Anything that would suffice it works.
How about something along these lines:
A first step to installing DALI is to install CUDA toolkit. There are many possible ways of installing it, e.g.:
1. [Using `pip install`](https://nvidia.github.io/cuda-python/install.html#installing-from-pypi): `pip install cuda-python`
1. [Using runfile](https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=22.04&target_type=runfile_local)
1. [Installing from conda](https://nvidia.github.io/cuda-python/install.html#installing-from-conda)
If the CUDA has been installed previously, it is not necessary to reinstall using any of these methods.
To install the latest DALI release [...]
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It is already in requirements, linked to the CUDA Toolkit installation guide https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html
which covers the topic extensively. I don't think it is good idea to copy a small subset of this page here.
Typically users working on DL will already have the CUDA installed for the purposes of the framework.
We don't specify how to install any of the frameworks or the driver either.
Signed-off-by: Krzysztof Lecki <klecki@nvidia.com>
Signed-off-by: Janusz Lisiecki <jlisiecki@nvidia.com>
Signed-off-by: Krzysztof Lecki <klecki@nvidia.com>
Signed-off-by: Krzysztof Lecki <klecki@nvidia.com>
!build |
CI MESSAGE: [7650371]: BUILD STARTED |
CI MESSAGE: [7650371]: BUILD PASSED |
Category: Other
Description:
Improve the documentation by linking to the CUDA toolkit installation instruction while mentioning it is required for CUDA 12-based builds.
Additional information:
Affected modules and functionalities:
docs/installation
Key points relevant for the review:
Better wording?
Tests:
Checklist
Documentation
DALI team only
Requirements
REQ IDs: N/A
JIRA TASK: N/A