-
Notifications
You must be signed in to change notification settings - Fork 43
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
support config gpu memory factor #28
Conversation
Signed-off-by: peiniliu <[email protected]>
| `--config-file` | `$CONFIG_FILE` | `""` | | ||
|
||
when starting volcano-device-plugin.yml, users can specify these parameters by adding args to the container 'volcano-device-plugin'. | ||
For example: | ||
- args: ["--gpu-strategy=number"] will let device plugin using the gpu-number strategy | ||
- args: ["--gpu-strategy=share","--gpu-memory-factor=10"] will let device plugin using the gpu-share strategy, and memory factor is 10MB |
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.
Are there some limit about the parameter gpu-memory-factor
? For example, could it be a decimal?
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.
This is not recommended generally, the most common setting is 1MiB / 1 GiB
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.
Yes, so I mean it should be highlight.
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.
/lgtm
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.
/lgtm
As described in #19 and #22, the device plugin needs to support config GPU memory factor to control the number of virtual devices that Listandwatched by kubelet. By default, we support each memory block as 1MB, this PR support config this factor, so that users can specify their own GPU memory granularity.
Tested Results:
2022/08/15 08:57:18 Loading configuration.
2022/08/15 08:57:18
Running with config:
{
"version": "v1beta1",
"flags": {
"GPUStrategy": "share",
"GPUMemoryFactor": 10
}
}
ENV of pod1.yml
NVIDIA_VISIBLE_DEVICES=0
NVIDIA_DRIVER_CAPABILITIES=compute,utility
VOLCANO_GPU_MEMORY_TOTAL=15109
VOLCANO_GPU_ALLOCATED=10240
Signed-off-by: peiniliu [email protected]