diff --git a/.github/workflows/docs.yml b/.github/workflows/docs.yml new file mode 100644 index 00000000..4e174cdb --- /dev/null +++ b/.github/workflows/docs.yml @@ -0,0 +1,54 @@ +name: Deploy Docs to GitHub Pages + +on: + push: + branches: [develop] + workflow_dispatch: + +permissions: + contents: read + pages: write + id-token: write + +concurrency: + group: pages + cancel-in-progress: false + +jobs: + build: + runs-on: ubuntu-22.04 + steps: + - uses: actions/checkout@v4 + + - uses: actions/setup-python@v5.4.0 + with: + python-version: "3.10" + + - name: Install dependencies + run: pip install -r docs/sphinx/requirements.txt + + - name: Build Sphinx docs + env: + DOCS_BASE_URL: https://amd-tme.github.io/instinct-docs/ + GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} + run: sphinx-build -b html docs/ docs/_build/html + + - name: Add .nojekyll + run: touch docs/_build/html/.nojekyll + + - uses: actions/configure-pages@v5 + + - uses: actions/upload-pages-artifact@v4 + with: + path: docs/_build/html + + deploy: + environment: + name: github-pages + url: ${{ steps.deployment.outputs.page_url }} + runs-on: ubuntu-22.04 + needs: build + steps: + - name: Deploy to GitHub Pages + id: deployment + uses: actions/deploy-pages@v4 diff --git a/.gitignore b/.gitignore index aaecec1c..5db81a7c 100644 --- a/.gitignore +++ b/.gitignore @@ -1,11 +1,13 @@ .venv .vscode build +.playwright-mcp/ # documentation artifacts _build/ _images/ _static/ +!docs/_static/ _templates/ _toc.yml docBin/ diff --git a/.wordlist.txt b/.wordlist.txt index 7d016eec..921a2b21 100644 --- a/.wordlist.txt +++ b/.wordlist.txt @@ -73,5 +73,6 @@ DeepSeek DriveNets minimap FFI +Slurm Slurm's WireGuard diff --git a/docs/_static/index.css b/docs/_static/index.css index 94250126..01209d10 100644 --- a/docs/_static/index.css +++ b/docs/_static/index.css @@ -8,8 +8,206 @@ max-width: 1440px !important; width: 100% !important; margin: 0 auto !important; - contain: layout; + contain: layout; } + +/* ─── Landing page only: give the card grid + "on this page" rail extra + room beyond the site's standard 1440px page width. The primary sidebar is + hidden by default here (see the .id-sidebar-toggle-input rules below), so + this 1432px cap is sized to match the main content's width in that default + state — the same effective width the card grid already had when the + sidebar used to always take its 20%-of-container cut out of a wider 1790px + box. If a visitor opts back into the sidebar, the grid shrinks to make + room for it rather than the page growing, which is an acceptable trade for + an opt-in control. Scoped via :has() on our own .id-landing-layout wrapper + so every other page is unaffected. ─── */ +.bd-main .bd-content .bd-article-container:has(.id-landing-layout) { + max-width: 1432px !important; +} +/* Upstream instinct-design.css also caps the nested
itself (and + .bd-page-width, further up the tree) at 1180px for any page with a hero — + both need the same landing-page escape hatch as .bd-article-container + above, or the grid stays starved regardless of the container override. */ +.bd-main .bd-content .bd-article-container:has(.id-landing-layout) > .bd-article { + max-width: 1432px !important; +} +.bd-page-width:has(.id-landing-layout) { + max-width: 1432px !important; +} + +/* The primary ("System Administrators" / "Common Reference" / ...) sidebar + duplicates the landing grid below it, so default it to collapsed here and + let visitors opt back in with the checkbox-driven toggle above. This can't + reuse the theme's own "Toggle primary sidebar" control (rocm_docs_theme's + toggle-primary-sidebar.html + custom.css) — that targets a checkbox + (input#__primary) from an older theme generation that no longer exists in + the pydata-sphinx-theme version this site runs, so the label is dead + markup on every page, not just this one. This is a plain checkbox hack + instead: the input lives inside .id-landing-main, and :has() reaches back + up to .bd-container (the nearest ancestor shared with .bd-sidebar-primary, + which lives in a separate branch of the DOM) to flip its display. */ +.id-sidebar-toggle-input { + position: absolute; + width: 1px; + height: 1px; + overflow: hidden; + opacity: 0; +} +.id-sidebar-toggle-label { + display: inline-flex; + align-items: center; + gap: 0.35rem; + font-size: 0.78rem; + font-weight: 600; + color: var(--id-muted-text); + cursor: pointer; + margin-bottom: 1rem; + user-select: none; +} +html[data-theme="light"] .id-sidebar-toggle-label { + color: var(--id-muted-text-light); +} +.id-sidebar-toggle-label:hover { + color: var(--id-teal); +} +html[data-theme="light"] .id-sidebar-toggle-label:hover { + color: var(--id-secondary); +} +.id-sidebar-toggle-label svg { + width: 18px; + height: 18px; + transition: transform 0.2s ease-in-out; +} +.id-sidebar-toggle-input:checked ~ .id-sidebar-toggle-label svg { + transform: rotate(90deg); +} +/* Desktop only (matches the theme's own $breakpoint-sidebar-primary: lg / + 960px, see pydata-sphinx-theme's _layout.scss) — below that width the + sidebar is already an off-canvas drawer opened via the theme's own mobile + hamburger toggle, which our checkbox has nothing to do with. Forcing + display:none unconditionally would fight that drawer and make it + impossible to open the sidebar at all on mobile. */ +@media (min-width: 960px) { + .bd-container:has(.id-landing-layout) .bd-sidebar-primary { + display: none; + } + .bd-container:has(.id-sidebar-toggle-input:checked) .bd-sidebar-primary { + display: flex; + } + /* rocm_docs_theme's own "Toggle primary sidebar" label (for="__primary") + duplicates our toggle here — it turns out not to be fully dead: a + pydata-sphinx-theme script grabs the first .primary-toggle element on + the page by class (regardless of the stale `for` target) and wires it + to the mobile off-canvas drawer's real checkbox. But that drawer only + has a visual effect below this same 960px breakpoint, so on desktop + clicking it does nothing — pure duplicate clutter next to our working + toggle. Hide it here only; leave it alone below 960px where it's still + the sole way to open the sidebar drawer on the landing page. */ + .bd-container:has(.id-landing-layout) label.primary-toggle { + display: none; + } +} +@media (max-width: 959.98px) { + .id-sidebar-toggle-label { + display: none; + } +} + +.id-landing-layout { + display: grid; + grid-template-columns: minmax(0, 1fr) 240px; + gap: 2.5rem; + align-items: start; +} +.id-landing-main { + min-width: 0; +} + +/* Plain vertical whitespace where the id-glow-line divider used to sit — + the divider itself is gone, but a bare margin-bottom on .id-hero-actions + read as too tight once it was removed. No visual mark, just breathing + room, roughly a

tag's worth. */ +.id-hero-spacer { + height: 10px; +} + +/* Deliberate 4-per-row card grid on the landing page at desktop widths. + The upstream default (auto-fill, minmax(248px, 1fr)) is what subpages + use, but here it would silently drop to 3 or 5 columns depending on + which deck a grid sits in (the tinted "Start Here" deck has its own + padding subpages don't), so pin it explicitly for a consistent look. */ +@media (min-width: 1000px) { + .id-landing-main .id-grid { + grid-template-columns: repeat(4, 1fr); + } +} + +/* "On this page" rail — reuses the same .bd-toc-nav/.page-toc classes and + markup shape as the interior-page Contents sidebar so it inherits the + site-wide Bootstrap scrollspy (data-bs-spy on ) and matches its + look, without depending on the theme's own (heading-driven) secondary + sidebar, which the landing page's raw-HTML sections don't populate. */ +.id-page-rail { + position: sticky; + top: calc(var(--pst-header-height, 4rem) + 1.5rem); + align-self: start; + max-height: calc(100vh - var(--pst-header-height, 4rem) - 3rem); + overflow-y: auto; + padding-top: 0.25rem; +} +.id-page-rail .page-toc.onthispage { + display: flex; + align-items: center; + gap: 0.5rem; + font-size: 0.7rem; + font-weight: 700; + letter-spacing: 0.1em; + text-transform: uppercase; + color: var(--id-muted-text); + margin-bottom: 0.75rem; +} +html[data-theme="light"] .id-page-rail .page-toc.onthispage { + color: var(--id-muted-text-light); +} +/* .bd-toc-nav never normally holds external links (the real Contents rail + is anchor-only), so the theme's external-link icon rule is scoped to + nav.bd-links and doesn't reach here — add it back for our one external + entry (Industries & Verticals) so it matches the sidebar's treatment. */ +.id-page-rail .bd-toc-nav a.reference.external::after { + content: var(--pst-icon-external-link); + font: var(--fa-font-solid); + font-size: 0.75em; + margin-left: 0.3em; +} + +.id-page-rail .bd-toc-nav a.active { + color: var(--id-teal); + font-weight: 600; +} +html[data-theme="light"] .id-page-rail .bd-toc-nav a.active { + color: var(--id-secondary); +} + +@media (max-width: 1200px) { + .id-landing-layout { + grid-template-columns: 1fr; + } + .id-page-rail { + display: none; + } +} + +/* Section names are plain links to their subpage, not a collapse toggle — + drop instinct-design.css's disclosure-triangle chevron so the link + doesn't still read as a dropdown, and suppress the browser's default + anchor underline (never needed when this was a

). */ +.id-subhead::before { + display: none; +} +.id-subhead { + text-decoration: none; +} + .small-sd-card-large.sd-card {} #buttonWrapper:hover { border-color: hsla(231, 99%, 66%, 1); @@ -20,10 +218,11 @@ h2 { margin: 0; font-size: 1.5em; } -.card-header { +.bd-main h2.card-header { font-size: 18px; font-weight: 700; line-height: 1.4; + margin-top: 0; margin-bottom: 10px; color: var(--text-color-light); display: -webkit-box; diff --git a/docs/conf.py b/docs/conf.py index ef87cae8..b21e4cd7 100644 --- a/docs/conf.py +++ b/docs/conf.py @@ -6,7 +6,10 @@ external_projects = ["gpu-operator", "device-metrics-exporter"] external_projects_path = "projects.yaml" -html_baseurl = os.environ.get("READTHEDOCS_CANONICAL_URL", "instinct.docs.amd.com") +html_baseurl = os.environ.get( + "READTHEDOCS_CANONICAL_URL", + os.environ.get("DOCS_BASE_URL", "instinct.docs.amd.com"), +) html_context = {} if os.environ.get("READTHEDOCS", "") == "True": html_context["READTHEDOCS"] = True @@ -22,8 +25,8 @@ html_copy_source = True html_theme = "rocm_docs_theme" html_theme_options = { - "flavor": "instinct", - "link_main_doc": False, + "flavor": "instinct-design", + "link_main_doc": True, "repository_url": "https://github.com/rocm/instinct-docs", "use_repository_button": True, "use_issues_button": True, @@ -41,12 +44,17 @@ html_static_path = ['_static', 'images'] +# Landing-page card layout. Loaded after the instinct-design flavor's +# own sheet so its card rules win on the index page. html_css_files = ["index.css"] # Table of contents external_toc_path = "./sphinx/_toc.yml" -exclude_patterns = ['.venv'] +# system-admin/_cards holds reusable `.id-card` HTML partials pulled in via +# MyST {include} directives (see docs/system-admin/*.md). They aren't +# standalone pages, so keep them out of document discovery / the toctree. +exclude_patterns = ['.venv', 'system-admin/_cards/*'] # Generate llms.txt and llms-full.txt after each build (the llms.txt standard, # https://llmstxt.org/). See the rocm-docs-core guide: @@ -83,5 +91,4 @@ def collapse_primary_sidebar(app, pagename, templatename, context, doctree): def setup(app): - app.add_css_file("css/index.css") app.connect("html-page-context", collapse_primary_sidebar) diff --git a/docs/images/mi3xx-reference.png b/docs/images/mi3xx-reference.png deleted file mode 100644 index c3076309..00000000 Binary files a/docs/images/mi3xx-reference.png and /dev/null differ diff --git a/docs/index.md b/docs/index.md index c62ee76b..8406191d 100644 --- a/docs/index.md +++ b/docs/index.md @@ -1,805 +1,264 @@ --- -html_theme.sidebar_secondary.remove: true +html_theme.sidebar_secondary.remove: true --- -# AMD Instinct Data Center GPU Documentation -The AMD Instinct Documentation site provides comprehensive guides and technical documentation for system administrators and technical users deploying AMD Instinct Data Center GPUs in enterprise environments. This site focuses on large-scale deployment, cluster management, monitoring, and operational best practices for both HPC and AI workloads. For API documentation and core software stack details, visit the [ROCm documentation](https://rocm.docs.amd.com). +
+
-::::::::::{dropdown} Industries/Verticals -:open: + + -::::::::{tab-set} +
+ AMD Instinct™ · Data Center GPU +
-:::::::{tab-item} Computer Vision +# Systems & Infrastructure Documentation -::::{grid} 2 2 3 4 - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/models.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Models and applications

- -

Design, develop, train, finetune, and infer your computer vision models on AMD Instinct GPUs. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/decoding.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Image and video decoding

- -

Decode a wide variety of image and video formats leveraging the speed of AMD Instinct GPUs. +

+Everything you need to deploy, validate, and operate AMD Instinct™ Data Center GPUs at scale — drivers, orchestration, cluster management, and acceptance testing for HPC and AI. For API and software-stack reference, see the ROCm documentation.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/processing.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Image processing

- -

Eliminate the image preprocessing bottleneck in your computer vision workloads with AMD Instinct GPUs. -

-::: - -:::: -::::::: +
+ + Install the GPU driver + + + + Customer Acceptance Guide + +
-:::::::{tab-item} Data Science - -::::{grid} 2 2 3 4 - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/ROCm-DS.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

ROCm-DS

- -

Accelerate your data science workloads with the AMD ROCm™ Data Science Toolkit (ROCm-DS). -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/hipDF.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

hipDF

- -

GPU Accelerated DataFrames to enhance your data manipulation and analysis tasks with AMD Instinct™ GPUs. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/hipGRAPH.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

hipGRAPH

- -

Create, manipulate, and analyze complex networks and graphs on AMD Instinct™ GPUs. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/hipVS.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

hipVS

- -

Accelerate a variety of vector search operations on AMD Instinct™ GPUs. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/hipMM.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

hipMM

- -

Advanced memory management utilities for the various ROCm-DS libraries. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/hipRAFT.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

hipRAFT

- -

Fundamental algorithms and primitives used by the various ROCm-DS libraries. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/ROCm-DS_Blogs.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Blogs

- -

Browse blogs detailing how to accelerate your data science workloads on AMD Instinct™ GPUs. -

-::: +
-:::: + -::::::: +
-:::::::{tab-item} Finance +
+
Start here
+

System Administrators

+

Deploy and run AMD Instinct GPUs on bare metal, in containers, and across clusters. These guides are the most frequently updated content on this site.

+
-::::{grid} 2 2 3 4 +
+Bare metal +
-:::{grid-item-card} -:padding: 1 -:img-top: ./images/finance-6.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

ROCm-Finance

- -

Boost your financial workloads with the ROCm Toolkit for Finance (ROCm-Finance). -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/finance-5.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

XGBoost

- -

General-purpose GPU gradient boosting. Start here for high-performance workloads for data-intensive applications. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/finance-10.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

LightGBM

- -

Leaf-wise training is the way to go. LightGBM is a strong fit when sparsity abounds and dataset size drives the bottleneck. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/finance-8.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

ThunderGBM

- -

GPU-oriented boosting for highly parallel, GPU-intensive training and simulation-style runs on high-dimensional datasets. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/finance-4.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

GitHub

- -

Source for all ROCm-Finance libraries on GitHub. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/finance-9.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Examples

+ + + Instinct GPU Driver + Install and configure the GPU, including logging and error codes. + -

Runnable examples on GitHub to explore the code. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/finance-7.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

ROCm-Finance Blogs

- -

Browse blogs detailing how to accelerate your finance workloads using gradient boosting on AMD Instinct GPUs. -

-::: - -:::: - -::::::: - -:::::::{tab-item} Life Science -::::{grid} 2 2 3 4 - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/ROCm-LS.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

ROCm-LS

- -

Accelerate your life science workloads with the ROCm Toolkit for Life Science (ROCm-LS). -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/hipCIM.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

hipCIM

- -

Run image processing and computer vision workloads on N-dimensional images for medical imaging workloads on AMD Instinct™ GPUs. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/MONAI.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

MONAI

- -

Train and deploy AI and computer vision models for medical imaging use cases on AMD Instinct™ GPUs. -

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/ROCm-LS_Blogs.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Blogs

+ + + GPU Partitioning + Split compute units and memory to partition a single GPU. + -

Browse blogs detailing how to accelerate your life science workloads on AMD Instinct™ GPUs. -

-::: - -:::: - -::::::: -:::::::{tab-item} Simulation & Modelling Apps -::::::{grid} 2 2 3 4 - -:::::{grid-item-card} -:padding: 1 -:img-top: ./images/AnsysFluent-tile.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Ansys Fluent

- -

Ansys Fluent is a computational fluid dynamics (CFD) software that leverages AMD GPU acceleration through ROCm for high-performance simulations. The AMD partnership enables exascale-ready CFD solutions for complex engineering challenges. -

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ./images/AnsysMech-tile.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Ansys Mechanical

+ + + AMD SMI + Unified user-space tool to manage and monitor GPUs and drivers. + -

Ansys Mechanical 2023R2+ supports MI200 -

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ./images/FCharLES-tile.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Cadence Fidelity LES Solver

- -

Fidelity LES Solver supports MI200. -

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ./images/DevitoPRO-tile.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Devito Codes DevitoPRO

- -

DevitoPRO 4.8.2+ supports MI200 and MI300X. -

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ./images/Starccm-tile.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Siemens Simcenter STAR-CCM+

- -

Simcenter STAR-CCM+ 2402+ supports MI200. -

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ./images/ECHELON-tile.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Stone Ridge Technology ECHELON

- -

ECHELON 2023.3+ supports MI200. -

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ./images/gsplat.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

GSplat

- -

Render 3D scenes and models with Gaussian splatting on MI300X. -

-::::: - -:::::: - -::::::: -:::::::::: - -::::::::::{dropdown} System Administrators -:open: - -::::::::{tab-set} - -:::::{tab-item} Bare Metal - -::::{grid} 2 2 3 4 - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/System-Administrators-Bare-Metal-AMD-GPU-Driver.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct GPU Driver

+ + + ROCm Validation Suite + System validation and hardware diagnostics. + -

Install and configure the GPU. Learn about logging including error codes.

-::: - -:::: - -::::: -:::::{tab-item} Containers and Orchestration Tools - -::::{grid} 2 2 3 4 - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/gpu-operator.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

GPU Operator

- -

The AMD GPU Operator simplifies the deployment and management of AMD Instinct GPU accelerators within Kubernetes clusters.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/network-operator.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Network Operator

- -

Network Operator simplifies the use of AMD AINICs in Kubernetes environments.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/device-plugin.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Device Plugin

- -

Kubernetes (k8s) device plugin to enable registration of AMD GPU to a container cluster

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/device-metrics-exporter.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Device Metrics Exporter

- -

The AMD Device Metrics Exporter enables Prometheus-format metrics collection for AMD GPUs in HPC and AI environments.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/container-toolkit.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD Container Toolkit

+ + + Customer Acceptance Guide + Configure, validate, benchmark, and baseline Instinct GPUs. + -

The AMD Container Toolkit enables seamless integration of AMD Instinct GPUs with Docker, offering streamlined GPU access, enhanced container runtime support, and optimized performance for ML and HPC workloads.

-::: -:::: - -::::: + + + Cluster Validation Suite + Test scripts that validate AMD AI clusters end to end. + + -:::::{tab-item} Cluster Management +
+
-::::{grid} 2 2 3 4 +
+Containers & orchestration +
-:::{grid-item-card} -:padding: 1 -:img-top: ./images/system-acceptance.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct Customer Acceptance Guide

- -

Comprehensive guide for configuring, validating, benchmarking, and baselining AMD Instinct™ GPUs.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/telemetry.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Cluster Networking Guide

- -

Optimize the network for Instinct GPU applications

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/mi3xx-reference.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

MI3XX Reference Design

+ + + GPU Operator + Deploy and manage Instinct GPUs in Kubernetes clusters. + -

Reference design materials and topology examples for the AMD Instinct MI3XX platform.

-::: -:::: + + + Network Operator + Simplify AMD AINICs in Kubernetes environments. + + -::::: + + + Device Plugin + Register AMD GPUs with a Kubernetes container cluster. + + -:::::{tab-item} Cloud + + + Device Metrics Exporter + Prometheus-format GPU metrics for HPC and AI environments. + + -::::{grid} 2 2 3 4 + + + AMD Container Toolkit + Integrate Instinct GPUs with Docker and container runtimes. + + -:::{grid-item-card} -:padding: 1 -:img-top: ./images/instinct-azure.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct on Azure

+ + + Spur + AI-native job scheduler, drop-in compatible with Slurm, with GPU-first scheduling and Raft-based state. + -

Get started with AMD Instinct on Azure

-::: -:::: +
+
-::::: +
+Cluster, cloud & virtualization +
-:::::{tab-item} Virtualization + + + Enterprise AI + Tools to manage enterprise AI infrastructure at scale. + + -::::{grid} 2 2 3 4 + + + Omnistat + Profile GPU resource utilization across the cluster. + + -:::{grid-item-card} -:padding: 1 -:img-top: ./images/virtualization_image.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct Virtualization Driver

+ + + Cluster Networking + Optimize the network for Instinct GPU applications. + -

Explore the Virtualization Driver for Instinct GPUs.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/instinct_virtualization_SMI.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD SMI Documentation

+ + + + Instinct on Azure + Get started with AMD Instinct GPUs on Microsoft Azure. + -

AMD SMI LIB enables you to manage and monitor AMD Virtualization Enabled GPUs.

-::: -:::: + + + Virtualization Driver + Explore the virtualization driver for Instinct GPUs. + + -::::: + + + AMD SMI for Virtualization + Manage and monitor virtualization-enabled AMD GPUs. + + -:::::::: +
+
-:::::::::: +
-::::::::::{dropdown} Resources -:open: + -::::::::{tab-set} +
-:::::{tab-item} Tools +
+

Common Reference

+

Architecture, programming models, and technical collateral that span every deployment.

+
-::::{grid} 2 2 3 4 +
-:::{grid-item-card} -:padding: 1 -:img-top: ./images/enterpriseAI.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Enterprise AI

+ + + Instinct Micro-architecture + Hardware details for MI350, MI300, MI200, and MI100 accelerators. + -

Tools to help enterprise customers address the growing need for AI infrastructure management.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/omnistat.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Omnistat

- -

Profile across the cluster.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/gpu-operator.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Cluster Validation Suite

- -

CVS is a collection of test scripts that validate AMD AI clusters.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/System-Administrators-Bare-Metal-AMD-AMD-SMI.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD SMI

- -

The AMD System Management Interface (AMD SMI) library offers a unified tool for managing and monitoring GPUs, particularly in high-performance computing environments. It provides a user-space interface that allows applications to control GPU operations, monitor performance, and retrieve information about the system’s drivers and GPUs.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/System-Administrators-Bare-Metal-AMD-Tools.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

ROCmValidationSuite

+ + + + AMD SMI API Reference + Full AMD SMI documentation covering all use cases. + -

System validation and diagnosis

-::: -:::: + + + HIP C++ + Learn the HIP programming model. + + -::::: + + + OpenMP + Explore the OpenMP programming model. + + -:::::{tab-item} Common Reference + + + Technical Information Portal + NDA technical documentation and design collateral. Login required. + + -::::{grid} 2 2 3 4 +
-:::{grid-item-card} -:padding: 1 -:img-top: ./images/instinct-microarchitecture.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct Micro-architecture

- -

Review hardware aspects of the AMD Instinct™ MI350, MI300, MI200 and MI100 series of GPU accelerators.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/System-Administrators-Bare-Metal-AMD-GPU-Partitioning.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

GPU Partitioning

- -

Learn how to split the compute units and memory to partition a GPU.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/hipcpp.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

HIP C++

- -

Learn about HIP programming model.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/virtualization_image.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

OpenMP

- -

Explore the OpenMP programming model.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/secure-docs.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD Technical Information Portal

- -

Additional NDA technical documentation, software, and design collateral for AMD Instinct products; -Login Required

-::: +
-:::: +
+ -::::: + -:::::::: +
+ -:::::::::: + diff --git a/docs/resources/common-reference.md b/docs/resources/common-reference.md deleted file mode 100644 index 5dd609e1..00000000 --- a/docs/resources/common-reference.md +++ /dev/null @@ -1,73 +0,0 @@ -# Common Reference - -::::::{grid} 2 2 3 4 - -:::::{grid-item-card} -:padding: 1 -:img-top: ../system-admin/images/System-Administrators-Bare-Metal-AMD-GPU-Partitioning.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

GPU Partitioning

- -

Learn how to split the compute units and memory to partition a GPU.

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ../images/instinct-microarchitecture.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct Micro-architecture

- -

Review hardware aspects of the AMD Instinct™ MI350, MI300, MI200 and MI100 series of GPU accelerators.

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ../images/hipcpp.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

HIP C++

- -

Learn about HIP programming model.

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ../images/virtualization_image.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

OpenMP

- -

Explore the OpenMP programming model.

-::::: - -:::::{grid-item-card} -:padding: 1 -:img-top: ../images/secure-docs.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD Technical Information Portal

- -

Additional NDA technical documentation, software, and design collateral for AMD Instinct products; -Login Required

-::::: - -:::::: - -:::::::::: diff --git a/docs/resources/index.md b/docs/resources/index.md deleted file mode 100644 index 6d0021cf..00000000 --- a/docs/resources/index.md +++ /dev/null @@ -1,72 +0,0 @@ -# Tools - -::::{grid} 2 2 3 4 - -:::{grid-item-card} -:padding: 1 -:img-top: ../system-admin/images/System-Administrators-Bare-Metal-AMD-AMD-SMI.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD SMI

- -

The AMD System Management Interface (AMD SMI) library offers a unified tool for managing and monitoring GPUs, particularly in high-performance computing environments. It provides a user-space interface that allows applications to control GPU operations, monitor performance, and retrieve information about the system’s drivers and GPUs.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ../system-admin/images/enterpriseAI.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Enterprise AI

- -

Tools for enterprise customers for AI infrastructure management.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ../system-admin/images/omnistat.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Omnistat

- -

Profile across the cluster.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ../system-admin/images/System-Administrators-Bare-Metal-AMD-Tools.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

ROCmValidationSuite

- -

System validation and diagnosis

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ../system-admin/images/gpu-operator.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Cluster Validation Suite

- -

CVS is a collection of test scripts that validate AMD AI clusters.

-::: - -:::: - -:::::::::: diff --git a/docs/sphinx/_toc.yml.in b/docs/sphinx/_toc.yml.in index 1d342fc1..f7298a56 100644 --- a/docs/sphinx/_toc.yml.in +++ b/docs/sphinx/_toc.yml.in @@ -1,70 +1,6 @@ root: index.md subtrees: -- caption: Industries/Verticals - entries: - - file: vision/index.md - title: Computer Vision - subtrees: - - entries: - - file: vision/ai.md - title: Models and applications - - file: vision/decode.md - title: Decoding - - file: vision/preprocess.md - title: Image processing - - file: data-science/index.md - title: Data Science (ROCm-DS) - subtrees: - - entries: - - file: data-science/hipDF.md - title: hipDF - - file: data-science/hipGRAPH.md - title: hipGRAPH - - file: data-science/hipVS.md - title: hipVS - - file: data-science/hipMM.md - title: hipMM - - file: data-science/hipRAFT.md - title: hipRAFT - - file: life-science/index.md - title: Life Science (ROCm-LS) - subtrees: - - entries: - - file: life-science/hipCIM.md - title: hipCIM - - entries: - - file: life-science/MONAI.md - title: MONAI - - file: finance/index.md - title: Finance (ROCm-Finance) - subtrees: - - entries: - - file: finance/xgboost.md - title: XGBoost - - file: finance/lightgbm.md - title: LightGBM - - file: finance/thundergbm.md - title: ThunderGBM - - file: isv-apps/index.md - title: Simulation & Modeling Apps - subtrees: - - entries: - - file: isv-apps/ansys-fluent.md - title: Fluent - - file: isv-apps/ansys-mechanical.md - title: Mechanical - - file: isv-apps/cadence-fidelity.md - title: Cadence Fidelity - - file: isv-apps/devito.md - title: DevitoPRO - - file: isv-apps/siemens.md - title: Siemens - - file: isv-apps/stone-ridge.md - title: Stone Ridge - - file: simulation/gsplat.md - title: GSplat - - caption: System Administrators entries: - file: system-admin/bare-metal.md @@ -73,6 +9,16 @@ subtrees: - entries: - url: https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/ title: AMD GPU Driver + - url: https://rocm.docs.amd.com/projects/amdsmi/en/latest/ + title: AMD-SMI + - url: https://rocm.docs.amd.com/projects/cvs/en/latest/index.html + title: Cluster Validation Suite + - url: https://rocm.docs.amd.com/projects/amdsmi/en/latest/conceptual/partition.html + title: GPU Partitioning + - url: https://instinct.docs.amd.com/projects/system-acceptance/en/latest/ + title: Instinct Customer Acceptance Guide + - url: https://rocm.docs.amd.com/projects/ROCmValidationSuite/en/latest/ + title: ROCm Validatation Suite - file: system-admin/co-tools.md title: Containers and Orchestration Tools subtrees: @@ -99,6 +45,10 @@ subtrees: title: Networking - url: https://instinct.docs.amd.com/projects/MI3XX-reference/latest/overview.html title: MI3XX Reference Design + - url: https://docs.amd.com/v/u/en-US/AMD_Instinct_AMD-DriveNets_System_Reference_Architecture_RF-72513 + title: AMD-DriveNets System Reference Architecture + - url: https://rocm.blogs.amd.com/artificial-intelligence/amd-comparative-analysis/README.html + title: Comparative Analysis of Scale-Out RoCE Network Traffic - file: system-admin/cloud.md title: Cloud subtrees: @@ -114,35 +64,12 @@ subtrees: - url: https://instinct.docs.amd.com/projects/amd-smi-virt/en/latest/ title: AMD SMI Documentation -- caption: Resources +# Industries/Verticals: no assigned owner yet and slated to move to its own +# repo, assumed for now to be gsid-verticals-docs (placeholder — update once +# a real destination is confirmed). Local content under vision/, data-science/, +# life-science/, finance/, isv-apps/ stays in this repo but is intentionally +# left out of the nav until that migration happens. +- caption: Industries/Verticals entries: - - file: resources/index.md - title: Tools - subtrees: - - entries: - - url: https://account.amd.com/en/forms/registration/enterpriseai-ea.html - title: Enterprise AI - - url: https://amdresearch.github.io/omnistat/ - title: Omnistat - - url: https://rocm.docs.amd.com/projects/cvs/en/latest/index.html - title: Cluster Validation Suite - - url: https://rocm.docs.amd.com/projects/amdsmi/en/latest/ - title: AMD SMI - - url: https://rocm.docs.amd.com/projects/ROCmValidationSuite/en/latest/ - title: ROCmValidationSuite - - file: resources/common-reference.md - title: Common Reference - subtrees: - - entries: - - url: https://rocm.blogs.amd.com/software-tools-optimization/compute-memory-modes/README.html - title: GPU Partitioning - - url: https://rocm.docs.amd.com/en/latest/reference/gpu-arch/index.html - title: Instinct Micro-architecture - - url: https://rocm.docs.amd.com/projects/amdsmi/en/latest/index.html - title: AMD SMI API Doc - - url: https://rocm.docs.amd.com/projects/HIP/en/latest/index.html - title: HIP C++ - - url: https://rocm.docs.amd.com/projects/llvm-project/en/latest/conceptual/openmp.html - title: OpenMP - - url: https://docs.amd.com/v/u/en-US/ug1729-amd-instinct-accelerators - title: AMD Technical Information Portal + - url: https://github.com/ROCm/gsid-verticals-docs + title: Industries & Verticals diff --git a/docs/system-admin/_cards/amd-container-toolkit.md b/docs/system-admin/_cards/amd-container-toolkit.md new file mode 100644 index 00000000..35663bfe --- /dev/null +++ b/docs/system-admin/_cards/amd-container-toolkit.md @@ -0,0 +1,6 @@ + + + AMD Container Toolkit + Integrate Instinct GPUs with Docker and container runtimes. + + diff --git a/docs/system-admin/_cards/amd-smi-virtualization.md b/docs/system-admin/_cards/amd-smi-virtualization.md new file mode 100644 index 00000000..7c6a9a4b --- /dev/null +++ b/docs/system-admin/_cards/amd-smi-virtualization.md @@ -0,0 +1,6 @@ + + + AMD SMI for Virtualization + Manage and monitor virtualization-enabled AMD GPUs. + + diff --git a/docs/system-admin/_cards/amd-smi.md b/docs/system-admin/_cards/amd-smi.md new file mode 100644 index 00000000..968b12b7 --- /dev/null +++ b/docs/system-admin/_cards/amd-smi.md @@ -0,0 +1,6 @@ + + + AMD SMI + Unified user-space tool to manage and monitor GPUs and drivers. + + diff --git a/docs/system-admin/_cards/cluster-networking.md b/docs/system-admin/_cards/cluster-networking.md new file mode 100644 index 00000000..bb0f19cc --- /dev/null +++ b/docs/system-admin/_cards/cluster-networking.md @@ -0,0 +1,6 @@ + + + Cluster Networking + Optimize the network for Instinct GPU applications. + + diff --git a/docs/system-admin/_cards/cluster-validation-suite.md b/docs/system-admin/_cards/cluster-validation-suite.md new file mode 100644 index 00000000..7a8e8b33 --- /dev/null +++ b/docs/system-admin/_cards/cluster-validation-suite.md @@ -0,0 +1,6 @@ + + + Cluster Validation Suite + Test scripts that validate AMD AI clusters end to end. + + diff --git a/docs/system-admin/_cards/customer-acceptance-guide.md b/docs/system-admin/_cards/customer-acceptance-guide.md new file mode 100644 index 00000000..5e3a0a5a --- /dev/null +++ b/docs/system-admin/_cards/customer-acceptance-guide.md @@ -0,0 +1,6 @@ + + + Customer Acceptance Guide + Configure, validate, benchmark, and baseline Instinct GPUs. + + diff --git a/docs/system-admin/_cards/device-metrics-exporter.md b/docs/system-admin/_cards/device-metrics-exporter.md new file mode 100644 index 00000000..6fd31e03 --- /dev/null +++ b/docs/system-admin/_cards/device-metrics-exporter.md @@ -0,0 +1,6 @@ + + + Device Metrics Exporter + Prometheus-format GPU metrics for HPC and AI environments. + + diff --git a/docs/system-admin/_cards/device-plugin.md b/docs/system-admin/_cards/device-plugin.md new file mode 100644 index 00000000..2631d453 --- /dev/null +++ b/docs/system-admin/_cards/device-plugin.md @@ -0,0 +1,6 @@ + + + Device Plugin + Register AMD GPUs with a Kubernetes container cluster. + + diff --git a/docs/system-admin/_cards/drivenets-reference.md b/docs/system-admin/_cards/drivenets-reference.md new file mode 100644 index 00000000..d8e6cf98 --- /dev/null +++ b/docs/system-admin/_cards/drivenets-reference.md @@ -0,0 +1,6 @@ + + + AMD-DriveNets System Reference Architecture + End-to-end reference architecture for large-scale AI GPU clusters, pairing AMD Instinct MI350-series (MI355X) compute with the DriveNets AI Fabric. + + diff --git a/docs/system-admin/_cards/enterprise-ai.md b/docs/system-admin/_cards/enterprise-ai.md new file mode 100644 index 00000000..fae50bdf --- /dev/null +++ b/docs/system-admin/_cards/enterprise-ai.md @@ -0,0 +1,6 @@ + + + Enterprise AI + Tools to manage enterprise AI infrastructure at scale. + + diff --git a/docs/system-admin/_cards/gpu-operator.md b/docs/system-admin/_cards/gpu-operator.md new file mode 100644 index 00000000..e3614689 --- /dev/null +++ b/docs/system-admin/_cards/gpu-operator.md @@ -0,0 +1,6 @@ + + + GPU Operator + Deploy and manage Instinct GPUs in Kubernetes clusters. + + diff --git a/docs/system-admin/_cards/gpu-partitioning.md b/docs/system-admin/_cards/gpu-partitioning.md new file mode 100644 index 00000000..ec6ab543 --- /dev/null +++ b/docs/system-admin/_cards/gpu-partitioning.md @@ -0,0 +1,6 @@ + + + GPU Partitioning + Split compute units and memory to partition a single GPU. + + diff --git a/docs/system-admin/_cards/instinct-gpu-driver.md b/docs/system-admin/_cards/instinct-gpu-driver.md new file mode 100644 index 00000000..4bbe4918 --- /dev/null +++ b/docs/system-admin/_cards/instinct-gpu-driver.md @@ -0,0 +1,6 @@ + + + Instinct GPU Driver + Install and configure the GPU, including logging and error codes. + + diff --git a/docs/system-admin/_cards/instinct-on-azure.md b/docs/system-admin/_cards/instinct-on-azure.md new file mode 100644 index 00000000..6382feae --- /dev/null +++ b/docs/system-admin/_cards/instinct-on-azure.md @@ -0,0 +1,6 @@ + + + Instinct on Azure + Get started with AMD Instinct GPUs on Microsoft Azure. + + diff --git a/docs/system-admin/_cards/mi3xx-reference.md b/docs/system-admin/_cards/mi3xx-reference.md new file mode 100644 index 00000000..fd8b16a6 --- /dev/null +++ b/docs/system-admin/_cards/mi3xx-reference.md @@ -0,0 +1,6 @@ + + + MI3XX Reference Design + Reference design materials and topology examples for the AMD Instinct MI3XX platform. + + diff --git a/docs/system-admin/_cards/network-operator.md b/docs/system-admin/_cards/network-operator.md new file mode 100644 index 00000000..6a9eff30 --- /dev/null +++ b/docs/system-admin/_cards/network-operator.md @@ -0,0 +1,6 @@ + + + Network Operator + Simplify AMD AINICs in Kubernetes environments. + + diff --git a/docs/system-admin/_cards/omnistat.md b/docs/system-admin/_cards/omnistat.md new file mode 100644 index 00000000..906b7618 --- /dev/null +++ b/docs/system-admin/_cards/omnistat.md @@ -0,0 +1,6 @@ + + + Omnistat + Profile GPU resource utilization across the cluster. + + diff --git a/docs/system-admin/_cards/roce-comparative-analysis.md b/docs/system-admin/_cards/roce-comparative-analysis.md new file mode 100644 index 00000000..37d995c5 --- /dev/null +++ b/docs/system-admin/_cards/roce-comparative-analysis.md @@ -0,0 +1,6 @@ + + + Comparative Analysis of Scale-Out RoCE Network Traffic + Compares RoCE network traffic patterns and loads across GPT-4, Llama 3, DeepSeek-V2, and Grok 4.0 LLM training. + + diff --git a/docs/system-admin/_cards/rocm-validation-suite.md b/docs/system-admin/_cards/rocm-validation-suite.md new file mode 100644 index 00000000..577b8dd3 --- /dev/null +++ b/docs/system-admin/_cards/rocm-validation-suite.md @@ -0,0 +1,6 @@ + + + ROCm Validation Suite + System validation and hardware diagnostics. + + diff --git a/docs/system-admin/_cards/spur.md b/docs/system-admin/_cards/spur.md new file mode 100644 index 00000000..e01a84dd --- /dev/null +++ b/docs/system-admin/_cards/spur.md @@ -0,0 +1,6 @@ + + + Spur + AI-native job scheduler, drop-in compatible with Slurm, with GPU-first scheduling and Raft-based state. + + diff --git a/docs/system-admin/_cards/virtualization-driver.md b/docs/system-admin/_cards/virtualization-driver.md new file mode 100644 index 00000000..2b86c13a --- /dev/null +++ b/docs/system-admin/_cards/virtualization-driver.md @@ -0,0 +1,6 @@ + + + Virtualization Driver + Explore the virtualization driver for Instinct GPUs. + + diff --git a/docs/system-admin/bare-metal.md b/docs/system-admin/bare-metal.md index f521fe35..0e33e4bb 100644 --- a/docs/system-admin/bare-metal.md +++ b/docs/system-admin/bare-metal.md @@ -1,18 +1,23 @@ -# Bare metal - -::::{grid} 2 2 2 2 - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/System-Administrators-Bare-Metal-AMD-GPU-Driver.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct GPU Driver

- -

Install and configure the GPU. Learn about logging including error codes.

-::: - -:::: +# Bare metal + +
+ +```{include} _cards/instinct-gpu-driver.md +``` + +```{include} _cards/gpu-partitioning.md +``` + +```{include} _cards/amd-smi.md +``` + +```{include} _cards/rocm-validation-suite.md +``` + +```{include} _cards/customer-acceptance-guide.md +``` + +```{include} _cards/cluster-validation-suite.md +``` + +
diff --git a/docs/system-admin/cloud.md b/docs/system-admin/cloud.md index 1db73caa..4decd8d3 100644 --- a/docs/system-admin/cloud.md +++ b/docs/system-admin/cloud.md @@ -1,18 +1,8 @@ -# Cloud - -::::{grid} 2 2 2 2 - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/instinct-azure.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct on Azure

- -

Get started with AMD Instinct on Azure

-::: - -:::: +# Cloud + +
+ +```{include} _cards/instinct-on-azure.md +``` + +
diff --git a/docs/system-admin/cluster.md b/docs/system-admin/cluster.md index f07ed125..1d5c0042 100644 --- a/docs/system-admin/cluster.md +++ b/docs/system-admin/cluster.md @@ -1,78 +1,32 @@ -# Cluster Documentation Hub +# Cluster -## Design and Guides +
-::::{grid} 2 4 4 4 +```{include} _cards/enterprise-ai.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/system-acceptance.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct Customer Acceptance Guide

- -

Comprehensive guide for configuring, validating, benchmarking, and baselining AMD Instinct™ GPUs.

-::: +```{include} _cards/omnistat.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/telemetry.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Cluster Networking Guide

- -

Optimize the network for Instinct GPU applications

-::: +```{include} _cards/cluster-networking.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/mi3xx-reference.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

MI3XX Reference Design

- -

Reference design materials and topology examples for the AMD Instinct MI3XX platform.

-::: +```{include} _cards/customer-acceptance-guide.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/amd-drivenets-system-reference.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD-DriveNets System Reference Architecture

- -

Validated, end-to-end reference architecture for building and operating large-scale AI GPU clusters using AMD Instinct MI350-series (MI355X) compute paired with the DriveNets AI Fabric networking solution.

-::: +```{include} _cards/mi3xx-reference.md +``` -:::: +```{include} _cards/drivenets-reference.md +``` -## Articles and Overviews +
-::::{grid} 2 4 4 4 +## Articles and overviews -:::{grid-item-card} -:padding: 1 -:img-top: ./images/RoCE-comparative-analysis.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Comparative Analysis of Scale-Out RoCE Network Traffic Patterns and Loads in Training Large Language Models

- -

Compares RoCE network traffic patterns and loads across GPT-4, Llama 3, DeepSeek-V2, and Grok 4.0 LLM training.

-::: +
-:::: +```{include} _cards/roce-comparative-analysis.md +``` + +
diff --git a/docs/system-admin/co-tools.md b/docs/system-admin/co-tools.md index 8337e0f7..f54c1542 100644 --- a/docs/system-admin/co-tools.md +++ b/docs/system-admin/co-tools.md @@ -1,83 +1,23 @@ # Containers and Orchestration Tools -::::{grid} 2 2 2 2 +
-:::{grid-item-card} -:padding: 1 -:img-top: ./images/gpu-operator.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

GPU Operator

- -

The AMD GPU Operator simplifies the deployment and management of AMD Instinct GPU accelerators within Kubernetes clusters.

-::: +```{include} _cards/gpu-operator.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/network-operator.png -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Network Operator

- -

Network Operator simplifies the use of AMD AINICs in Kubernetes environments.

-::: +```{include} _cards/network-operator.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/device-plugin.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Device Plugin

- -

Kubernetes (k8s) device plugin to enable registration of AMD GPU to a container cluster

-::: +```{include} _cards/device-plugin.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/device-metrics-exporter.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Device Metrics Exporter

- -

The AMD Device Metrics Exporter enables Prometheus-format metrics collection for AMD GPUs in HPC and AI environments.

-::: +```{include} _cards/device-metrics-exporter.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/container-toolkit.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD Container Toolkit

- -

The AMD Container Toolkit enables seamless integration of AMD Instinct GPUs with Docker, offering streamlined GPU access, enhanced container runtime support, and optimized performance for ML and HPC workloads.

-::: +```{include} _cards/amd-container-toolkit.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/telemetry.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Spur

- -

An AI-native job scheduler written in Rust. Drop-in compatible with Slurm's CLI, REST API, and C FFI — with WireGuard mesh networking, GPU-first scheduling, and Raft-based state that survives restarts.

-::: +```{include} _cards/spur.md +``` -:::: +
diff --git a/docs/system-admin/images/RoCE-comparative-analysis.png b/docs/system-admin/images/RoCE-comparative-analysis.png deleted file mode 100644 index c0aa037b..00000000 Binary files a/docs/system-admin/images/RoCE-comparative-analysis.png and /dev/null differ diff --git a/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-AMD-SMI.jpg b/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-AMD-SMI.jpg deleted file mode 100644 index de88412d..00000000 Binary files a/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-AMD-SMI.jpg and /dev/null differ diff --git a/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-GPU-Driver.jpg b/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-GPU-Driver.jpg deleted file mode 100644 index a8661ebb..00000000 Binary files a/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-GPU-Driver.jpg and /dev/null differ diff --git a/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-GPU-Partitioning.jpg b/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-GPU-Partitioning.jpg deleted file mode 100644 index dcae1073..00000000 Binary files a/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-GPU-Partitioning.jpg and /dev/null differ diff --git a/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-Tools.jpg b/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-Tools.jpg deleted file mode 100644 index 6b4bea52..00000000 Binary files a/docs/system-admin/images/System-Administrators-Bare-Metal-AMD-Tools.jpg and /dev/null differ diff --git a/docs/system-admin/images/amd-drivenets-system-reference.png b/docs/system-admin/images/amd-drivenets-system-reference.png deleted file mode 100644 index 0ba0ebc9..00000000 Binary files a/docs/system-admin/images/amd-drivenets-system-reference.png and /dev/null differ diff --git a/docs/system-admin/images/container-toolkit.jpg b/docs/system-admin/images/container-toolkit.jpg deleted file mode 100644 index eebd4dc9..00000000 Binary files a/docs/system-admin/images/container-toolkit.jpg and /dev/null differ diff --git a/docs/system-admin/images/device-metrics-exporter.jpg b/docs/system-admin/images/device-metrics-exporter.jpg deleted file mode 100644 index 85de1d9a..00000000 Binary files a/docs/system-admin/images/device-metrics-exporter.jpg and /dev/null differ diff --git a/docs/system-admin/images/device-plugin.jpg b/docs/system-admin/images/device-plugin.jpg deleted file mode 100644 index 23e9034b..00000000 Binary files a/docs/system-admin/images/device-plugin.jpg and /dev/null differ diff --git a/docs/system-admin/images/enterpriseAI.png b/docs/system-admin/images/enterpriseAI.png deleted file mode 100644 index 1d60528c..00000000 Binary files a/docs/system-admin/images/enterpriseAI.png and /dev/null differ diff --git a/docs/system-admin/images/gpu-operator.jpg b/docs/system-admin/images/gpu-operator.jpg deleted file mode 100644 index 9658d13a..00000000 Binary files a/docs/system-admin/images/gpu-operator.jpg and /dev/null differ diff --git a/docs/system-admin/images/instinct-azure.jpg b/docs/system-admin/images/instinct-azure.jpg deleted file mode 100644 index 788ae1d7..00000000 Binary files a/docs/system-admin/images/instinct-azure.jpg and /dev/null differ diff --git a/docs/system-admin/images/instinct_virtualization.jpg b/docs/system-admin/images/instinct_virtualization.jpg deleted file mode 100644 index 93f1cce8..00000000 Binary files a/docs/system-admin/images/instinct_virtualization.jpg and /dev/null differ diff --git a/docs/system-admin/images/instinct_virtualization_SMI.jpg b/docs/system-admin/images/instinct_virtualization_SMI.jpg deleted file mode 100644 index a1077359..00000000 Binary files a/docs/system-admin/images/instinct_virtualization_SMI.jpg and /dev/null differ diff --git a/docs/system-admin/images/mi3xx-reference.png b/docs/system-admin/images/mi3xx-reference.png deleted file mode 100644 index c3076309..00000000 Binary files a/docs/system-admin/images/mi3xx-reference.png and /dev/null differ diff --git a/docs/system-admin/images/network-operator.png b/docs/system-admin/images/network-operator.png deleted file mode 100644 index c3d87615..00000000 Binary files a/docs/system-admin/images/network-operator.png and /dev/null differ diff --git a/docs/system-admin/images/omnistat.jpg b/docs/system-admin/images/omnistat.jpg deleted file mode 100644 index 0dfccfce..00000000 Binary files a/docs/system-admin/images/omnistat.jpg and /dev/null differ diff --git a/docs/system-admin/images/system-acceptance.jpg b/docs/system-admin/images/system-acceptance.jpg deleted file mode 100644 index 84bdc2da..00000000 Binary files a/docs/system-admin/images/system-acceptance.jpg and /dev/null differ diff --git a/docs/system-admin/images/telemetry.jpg b/docs/system-admin/images/telemetry.jpg deleted file mode 100644 index 0a5035f4..00000000 Binary files a/docs/system-admin/images/telemetry.jpg and /dev/null differ diff --git a/docs/system-admin/index.md b/docs/system-admin/index.md index ee2fb474..acb32b2c 100644 --- a/docs/system-admin/index.md +++ b/docs/system-admin/index.md @@ -1,109 +1,75 @@ -# System Administrators - -::::{grid} 2 2 2 2 - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/System-Administrators-Bare-Metal-AMD-GPU-Driver.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct GPU Driver

- -

Install and configure the GPU. Learn about logging including error codes.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/system-acceptance.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

MI300X System Acceptance Tests

- -

Test the correct functioning and optimal performance of server systems equipped with AMD Instinct MI300X GPU accelerators.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/gpu-operator.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

GPU Operator

- -

The AMD GPU Operator simplifies the deployment and management of AMD Instinct GPU accelerators within Kubernetes clusters.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/device-plugin.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Device Plugin

- -

Kubernetes (k8s) device plugin to enable registration of AMD GPU to a container cluster

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/device-metrics-exporter.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Device Metrics Exporter

- -

The AMD Device Metrics Exporter enables Prometheus-format metrics collection for AMD GPUs in HPC and AI environments.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/container-toolkit.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD Container Toolkit

- -

The AMD Container Toolkit enables seamless integration of AMD Instinct GPUs with Docker, offering streamlined GPU access, enhanced container runtime support, and optimized performance for ML and HPC workloads.

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/telemetry.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Cluster Networking

- -

Optimize the network for Instinct GPU applications

-::: - -:::{grid-item-card} -:padding: 1 -:img-top: ./images/instinct-azure.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct on Azure

- -

Get started with AMD Instinct on Azure

-::: - -:::: +# System Administrators + +Deploy and run AMD Instinct GPUs on bare metal, in containers, and across clusters. + +## Bare metal + +
+ +```{include} _cards/instinct-gpu-driver.md +``` + +```{include} _cards/gpu-partitioning.md +``` + +```{include} _cards/amd-smi.md +``` + +```{include} _cards/rocm-validation-suite.md +``` + +```{include} _cards/customer-acceptance-guide.md +``` + +```{include} _cards/cluster-validation-suite.md +``` + +
+ +## Containers & orchestration + +
+ +```{include} _cards/gpu-operator.md +``` + +```{include} _cards/network-operator.md +``` + +```{include} _cards/device-plugin.md +``` + +```{include} _cards/device-metrics-exporter.md +``` + +```{include} _cards/amd-container-toolkit.md +``` + +```{include} _cards/spur.md +``` + +
+ +## Cluster, cloud & virtualization + +
+ +```{include} _cards/enterprise-ai.md +``` + +```{include} _cards/omnistat.md +``` + +```{include} _cards/cluster-networking.md +``` + +```{include} _cards/instinct-on-azure.md +``` + +```{include} _cards/virtualization-driver.md +``` + +```{include} _cards/amd-smi-virtualization.md +``` + +
diff --git a/docs/system-admin/virtualization.md b/docs/system-admin/virtualization.md index c716c0a1..7110ae78 100644 --- a/docs/system-admin/virtualization.md +++ b/docs/system-admin/virtualization.md @@ -1,31 +1,11 @@ # Virtualization -::::{grid} 2 2 3 4 +
-:::{grid-item-card} -:padding: 1 -:img-top: ./images/instinct_virtualization.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

Instinct Virtualization Driver

- -

Explore the Virtualization Driver for Instinct GPUs.

-::: +```{include} _cards/virtualization-driver.md +``` -:::{grid-item-card} -:padding: 1 -:img-top: ./images/instinct_virtualization_SMI.jpg -:class-img-top: small-sd-card-img-top -:class-body: small-sd-card -:class: small-sd-card -+++ - -

AMD SMI Documentation

- -

AMD SMI LIB enables you to manage and monitor AMD Virtualization Enabled GPUs.

-::: +```{include} _cards/amd-smi-virtualization.md +``` -:::: +