chore: initial k8s rocm support [CM-367] #9794
Conversation
✅ Deploy Preview for determined-ui ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
NicholasBlaskey
force-pushed
the
k8s_rocm
branch
from
35bc0df to
763a121
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #9794 +/- ##
==========================================
- Coverage 54.38% 54.01% -0.38%
==========================================
Files 1261 1261
Lines 155770 155795 +25
Branches 3540 3539 -1
==========================================
- Hits 84711 84146 -565
- Misses 70921 71511 +590
Partials 138 138
Flags with carried forward coverage won't be shown. Click here to find out more.
|
NicholasBlaskey
changed the title
| @@ -403,14 +403,21 @@ resource pool ``max_slots_per_pod``. | |||
| ``slot_type`` | |||
| ------------- | |||
|
|
|||
| Resource type used for compute tasks. Defaults to ``cuda``. | |||
| Resource type used for compute tasks. Valid options are ``gpu``, ``cuda``, ``cpu``, or ``rocm``. | |||
| Defaults to ``cuda``. | |||
There was a problem hiding this comment.
Does it default to cuda or gpu? The helm config reference says gpu?
There was a problem hiding this comment.
The master config defaults to cuda, helm defaults to gpu
docs/release-notes/rocm-k8s-gpu.rst
Outdated
| **New Features** | ||
|
|
||
| - Kubernetes: Experimental support for AMD ROCM GPUs is now available for Kubernetes. To use set | ||
| ``slotType=rocm``. See :ref:`helm-config-reference` for more details. |
There was a problem hiding this comment.
nit: I think the docs style guideline prefers visit over see.
| case device.CPU, device.CUDA: | ||
| case device.ROCM: | ||
| checkSlotType = errors.Errorf("rocm slot_type is not supported yet on k8s") | ||
| case device.CPU, device.CUDA, device.ROCM: |
There was a problem hiding this comment.
Does the code detect when the user inputs gpu and switches it to cuda? This makes it seem like gpu isn't actually accepted for slotType?
There was a problem hiding this comment.
Yeah we switch it here
NicholasBlaskey
force-pushed
the
k8s_rocm
branch
from
d00685d to
0a054f1
Compare
Ticket
Description
Adds initial experimental support for amd gpus on k8s.
Test Plan
Unit tests cover this change
Also manually verified on amd hardware. Automated e2e tests were deemed not in scope due to hardware availability issues
blocked on determined-ai/environments#275. We just need the environment images to land in the same release as this. Some additional docs will be added in the environments pr detailing what our rocm environment does and does not support.
Checklist
docs/release-notes/See Release Note for details.