[DRAFT] - ENG-4292 - Initial changes for ODH AMD GPU docs #561
Conversation
chtyler
force-pushed
the
ENG-4392-amd-gpus-docs
branch
from
06044ee to
7355f4b
Compare
chtyler
force-pushed
the
ENG-4392-amd-gpus-docs
branch
from
7355f4b to
844091f
Compare
…everal modules and correct some formatting issues and content errors
chtyler
force-pushed
the
ENG-4392-amd-gpus-docs
branch
from
5264ee9 to
bd012fa
Compare
chtyler
changed the title
| ifndef::upstream[] | ||
| | Ray ROCm | ||
| endif::[] | ||
| ifdef::upstream[] | ||
| | Ray ROCm | ||
| | If you are working with compute-intensive models and you want to accelerate the training job with AMD GPU support, you can use the Ray ROCm base image to gain access to the AMD ROCm software stack. | ||
| Using this software stack, you can accelerate your work by using libraries and tools that are optimized for AMD GPUs. |
There was a problem hiding this comment.
Is this missing an endif::[] here?
|
|
||
| You can use AMD GPUs with {productname-short} to accelerate AI and machine learning (ML) workloads. AMD GPUs provide high-performance compute capabilities, allowing users to process large data sets, train deep neural networks, and perform complex inference tasks more efficiently. | ||
|
|
||
| Integrating AMD GPUs with {productname-short} involves the following components and benefits: |
There was a problem hiding this comment.
Suggestion (from comment on line 16):
Integrating AMD GPUs with {productname-short} involves the following components:
| * **AMD GPU Operator**: | ||
| The AMD GPU Operator simplifies GPU integration by automating driver installation, device plugin setup, and node labeling for GPU resource management. It ensures compatibility between OpenShift and AMD hardware while enabling scaling of GPU-enabled workloads. | ||
|
|
||
| * **Why Use AMD GPUs with {productname-short}?** |
There was a problem hiding this comment.
This third bullet doesn't really flow with the first two. Suggestion:
Integrating AMD GPUs with {productname-short} provides the following benefits:
There was a problem hiding this comment.
I'd delete this whole section, the intro paragraph ("AMD GPUs provide high-performance compute capabilities, [...]") is imo sufficient.
| * Your OpenShift environment supports EC2 DL1 instances if you are running on Amazon Web Services (AWS). | ||
| endif::[] | ||
| ifdef::cloud-service[] | ||
| * You have logged in to OpenShift. | ||
| * You have the `cluster-admin` role in OpenShift. | ||
| * You have installed your AMD GPU and confirmed that it is detected in your environment. | ||
| * Your OpenShift environment supports EC2 DL1 instances if you are running on Amazon Web Services (AWS). |
There was a problem hiding this comment.
Could use {openshift-platform} instead of OpenShift
| endif::[] | ||
|
|
||
| .Procedure | ||
| . Install version 1.1.1 of the AMD GPU Operator, as described in link:https://dcgpu.docs.amd.com/projects/gpu-operator/en/main/installation/openshift-olm.html[Install AMD GPU Operator on OpenShift]. |
There was a problem hiding this comment.
Is it important to specify the version here? If not, it might be easier to leave it out to avoid having to update it later. Or create an attribute.
| ** To enable NVIDIA GPUs on OpenShift, you must install the link:https://docs.nvidia.com/datacenter/cloud-native/openshift/latest/index.html[NVIDIA GPU Operator]. | ||
| * AMD graphics processing units (GPUs) | ||
| ** Use the AMD GPU Operator to enable AMD GPUs for workloads such as AI/ML training and inference. | ||
| ** To enable AMD GPUs on OpenShift, you must: |
There was a problem hiding this comment.
** To enable AMD GPUs on {openshift-platform}, you must do the following tasks:
| * You have a running OpenShift cluster with a node equipped with an AMD GPU. | ||
| * You have access to the OpenShift CLI (`oc`) and terminal access to the node. | ||
|
|
||
| .Procedure | ||
| . Use the OpenShift CLI to verify if GPU resources are allocatable: |
There was a problem hiding this comment.
Could use {openshift-platform} instead of OpenShift
| ---- | ||
| oc describe node <node_name> | ||
| ---- | ||
| .. In the output, locate the **Capacity** and **Allocatable** sections and confirm that `amd.com/gpu` is listed. Example output: |
There was a problem hiding this comment.
Suggestion: For example: instead of Example output: for consistency.
| ---- | ||
| lspci | grep -E "MI210|MI250|MI300" | ||
| ---- | ||
| .. Verify that the output includes one of the AMD GPU models. Example: |
There was a problem hiding this comment.
Suggestion: Use For example: instead of Example:
|
|
||
| .Verification | ||
| * The `oc describe node <node_name>` command lists `amd.com/gpu` under **Capacity** and **Allocatable**. | ||
| * The `lspci` command output identifies an AMD GPU as a PCI device matching one of the specified models (e.g., MI210, MI250, MI300). |
There was a problem hiding this comment.
Per Style Guide, do not use e.g - replace with:
(for example, MI210, MI250, MI300)
| * AMD graphics processing units (GPUs) | ||
| ** Use the AMD GPU Operator to enable AMD GPUs for workloads such as AI/ML training and inference. | ||
| ** To enable AMD GPUs on OpenShift, you must: | ||
| *** Install version 1.1.1 of the AMD GPU Operator. |
There was a problem hiding this comment.
Is it important to specify the version here? If not, it might be easier to leave it out to avoid having to update it later. Or create an attribute.
| * AMD graphics processing units (GPUs) | ||
| ** Use the AMD GPU Operator to enable AMD GPUs for workloads such as AI/ML training and inference. | ||
| ** To enable AMD GPUs on OpenShift, you must: | ||
| *** Install version 1.1.1 of the AMD GPU Operator. |
chtyler
changed the title
Description
These are the changes to document the introduction of the AMD GPUs GA at 2.16.1 for ODH. AMD have provided all of the in-depth enablement instructions for deploying the AMD GPU Operator on RHOAI. Therefore, we link out to the AMD instructions where and when possible. In some areas, we have needed to add our own content.
How Has This Been Tested?
Merge criteria: