-
Notifications
You must be signed in to change notification settings - Fork 1
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 explicit GPU models for all makes #111
base: master
Are you sure you want to change the base?
Conversation
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.
Looks good, lgtm once comments addressed
@@ -121,8 +121,12 @@ class NodePool: | |||
disk_size: int | None = None | |||
disk_type: str | None = None | |||
|
|||
gpu: int | None = None |
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.
You cannot just drop these fields, it breaks SDK/CLI.
Some migration procedure is required.
I suggest keeping gpu
/ gpu_model
here for a while (say, until the New Year).
gpu could be a sum of amd, intel and nvidia.
gpu_model could be a first available model or concatination of available modes. We don't parse these values, so nvidia-tesla-k80+amd-rizen-xxx
would be fine.
Later, when all clients update CLI/FLOW on their machines, we can drop these fields.
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.
@asvetlov I've restored both fields for now to not break the compatibility with the SDK.
Just FYI:
gpu
was already dropped from the platform config api at April: https://github.com/neuro-inc/platform-config/commit/fc83088224859d5e81b35beb426069888e05c336#diff-61fdfd6f40520f332212d51ba625421a707ce6ff111f58610dd10c84b27b1d9b
If gpu
property is sent to the API, it's converted by the API to an nvidia_gpu
under the hood - https://github.com/neuro-inc/platform-config/commit/fc83088224859d5e81b35beb426069888e05c336#diff-c2827afbcb2bed90d9593131e0055607f29b65656938e4b93d972c2d40c7e27eR118
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.
I assume that gpu
property is actually blank in a DB since April? @zubenkoivan can you elaborate please?
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.
In April we renamed gpu to nvidia-gpu and added two other gpu kinds. Because before the April we didn't have non-nvidia systems.
I like your approach for keeping gpu
as an alias for nvidia, e.g. nvidia_gpu=payload.get("nvidia_gpu") or payload.get("gpu"),
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.
Should NodePoolOptions
and Resource
have triple nvidia/amd/intel values like NodePool
?
If not please elaborate why.
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.
@asvetlov
Resource
is used for idle jobs only. I've asked @zubenkoivan about that and he mentioned that it is not required to have a different manufacturers support there.
NodePoolOptions
are used only be a cloud providers. As far as I understand - we don't need to add such support for clouds.
@zubenkoivan please correct me if I'm wrong.
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.
correct, we don't need to support this for NodePoolOptions and Resource classes
d7cf539
to
7ba9d6d
Compare
for more information, see https://pre-commit.ci
No description provided.