Repository navigation
docs: copyedit v0.3.0 content and fix install driver opener - #957
Conversation
Light editorial pass over prose added since v0.2.1 in README.md, docs/vllm.md, and docs/wsl.md: em dashes, "rather than", "e.g.", "vs.", British spellings, and unwrapped paragraphs in the new WSL serve section. The one content fix is the `install driver` opener in README.md, which now says the command previews by default and that `--dkms` and `--yes` select and apply the DKMS plan. No other new facts. Backport note for v0.3.0: copyedits and one sentence rewrite in the README, vllm.md, and wsl.md sources that the site includes. No include anchors or wrapper files changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Follow-up copyedit after a self-review of the previous commit: remove the ambiguous "the second way" in the WSL librocdxg paragraph, split a run-on in the serve-reuse note, change "points at" to "points to", split the long OOM-hint sentence, and capitalize "Unix". The `--dkms`/`--yes` and `--verbose`/`--foreground` sentences are reworded for clarity. No facts changed. Backport note for v0.3.0: wording-only edits in the README, vllm.md, and wsl.md sources that the site includes. No include anchors or wrapper files changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
The standalone "See ROCm CLI in action" page was two sentences and two GIFs. Show the ROCm CLI recording on the Getting started page after the first-run steps, and remove the Demos page, its table of contents entry, and its link on the landing page. The README "Demos" section is not changed, so it still renders on GitHub. Backport note for v0.3.0: site wrappers only (getting-started.md, index.rst, _toc.yml.in, and the removed demos.md). No source-file prose or include anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Add subheadings to the Model serving section and turn dense paragraphs (--wait, --gpu, the reuse conditions, the --engine-env secret rules, and the ComfyUI reinstall behaviors) into lead-ins with bullets. Split the --port/--host, AIM-options, and lenient-model-name paragraphs. Only sentence boundaries changed, plus one lead-in sentence; no facts, flags, or include anchors changed. Backport note for v0.3.0: docs only, safe to cherry-pick onto the release branch. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
5bcd122 to
d9aaf5c
Compare
Convert the site landing page from index.rst to index.md, matching the other pages. Replace the three category cards (Install, Getting started, Use ROCm CLI) with four task cards that link to the same pages, with WSL2 as a second link on the Install card. Card text is taken from what those pages already say. Flatten the left nav from four captions to three: Get started (install, WSL2, getting started), Reference (command reference, vLLM adapter), and About. No pages added, renamed, or removed, so no redirects are needed. Backport note for v0.3.0: docs only. Depends on the docs/0.3.0 changes, which already edit the landing page and nav, so apply it after them. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
The repository front page opens with the CLI demo after the intro and Tech Preview note. Show the same recording on the site landing page, with the caption sentence from the README. The Getting started page keeps its copy next to the steps it shows. Backport note for v0.3.0: docs only. Applies on top of the landing page and nav commit. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
…ader Add a Release notes page as the first table of contents entry, written from the v0.3.0-rc.1 release notes in the team release-notes format: release highlights, breaking changes, added, changed, resolved issues, and known issues. Issue numbers are dropped and the GitHub releases page is linked. Add a release history page (0.3.0 and 0.2.1) reached from a Version List link in the header. Set the version in conf.py so the header and the left nav title show ROCm CLI 0.3.0, and point the ROCm Core SDK link at latest because that project has no docs-0.3.0 build. The 0.3.0 release date is listed as "To be announced" until the final release. Backport note for v0.3.0: cherry-pick this commit onto the v0.3.0 docs branch; it only touches docs/rocm-docs. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Group the resolved issues by area, replace implementation wording with what the user sees, define rocmd and the Doctor skill on first use, merge the vLLM startup diagnostics into one bullet, and trim the release highlights so they do not repeat the Added section. Backport note for v0.3.0: cherry-pick this commit with the release notes commit; it only touches docs/rocm-docs/release-notes.md. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Reword the transactional runtime, port claiming, and highlight bullets in terms of what the user sees. No facts added or removed. Backport note for v0.3.0: cherry-pick this commit with the release notes commits; it only touches docs/rocm-docs/release-notes.md. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Name the fix ID for the Visual C++ runtime check instead of the generic "print-only fix", and state what `rocm fix` does with it, as the README describes. Backport note for v0.3.0: cherry-pick onto release/v0.3.0 with the other docs/0.3.0 commits. It only touches docs/rocm-docs/release-notes.md. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
The same recording is on the landing page. Remove it and its lead-in sentence from Getting started so the page does not repeat it. Backport note for v0.3.0: cherry-pick onto release/v0.3.0 with the other docs/0.3.0 commits. It only touches docs/rocm-docs/getting-started.md. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
…ables Split one long Doctor sentence, say that the Doctor skill requires the failure catalog, and simplify one Changed bullet. Add a Breaking changes bullet for the retired ROCM_CLI_AMDGPU_VERSION and ROCM_CLI_AMDGPU_PACKAGE_RELEASE variables, taken from the README driver section. Backport note for v0.3.0: cherry-pick onto release/v0.3.0 with the other docs/0.3.0 commits. It only touches docs/rocm-docs/release-notes.md. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Remove "real" and "simply", expand "GPU arch" to "GPU architecture", replace "since" with "because" and "may" with "might" for possibility, replace "and so on" with a concrete example, and use code font for the Ctrl-D and Ctrl-C keys to match the rest of the page. Backport note for v0.3.0: wording only, no facts or anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Add a language to the two bare code fences, replace "rather than" and "since" with "instead of" and "because", use "might" for possibility, and expand IPC on first use. Backport note for v0.3.0: wording only, no facts or anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Replace "rather than" with "instead of", turn the "Source-build alternative." fragment into a full sentence, and use "can" instead of "may" for the example version that differs per host. Backport note for v0.3.0: wording only, no facts or anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Use "instead of" in the out-of-memory note description. Backport note for v0.3.0: wording only, no facts or anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Split the glibc requirement paragraph and name the Lemonade engine instead of "it", make the "command not found" tip shell-neutral, and break the quit-confirmation paragraph into shorter paragraphs with a list for the prompt choices. Backport note for v0.3.0: wording and structure only, no facts or anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Break the pre-push, license-headers, and module-organization paragraphs into shorter ones, turn the workspace-wide triggers into a list, add lead-in sentences before the tables and lists, and link the repository files cited in text with absolute GitHub URLs so they resolve on the docs site. Backport note for v0.3.0: wording and structure only, no facts or anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Move the Ubuntu 22.04 detail out of the prerequisites list into its own paragraph, and start a new paragraph for the ROCDXG install plan. Backport note for v0.3.0: wording and structure only, no facts or anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Add the missing "First run" heading above the first included section and place "Interactive interfaces" before "Configure ROCm and serve a model" to match the README order. Backport note for v0.3.0: structure only, no facts or anchors changed. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Convert the diagnose --report, serve --follow, and OOM control-byte paragraphs into lists and shorter paragraphs, split long sentences in the runtime, cleanup, ComfyUI, and bench sections, expand VRAM, OOM, NDJSON, and TLS on first use, put the remaining Ctrl-C mentions in code font, add a serial comma to the gfx list, and give each shell its own block in the completions section. Backport note for v0.3.0: documentation-only change to README.md, which the published site includes. Safe to cherry-pick to release/v0.3.0. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
The install page now says ROCm CLI ships prebuilt binaries (rocm and rocmd), so the README, landing page, and Install card no longer say "single prebuilt binary", and the banner reads "one install". The demo GIFs now sit on the Getting started page so console.gif stays linked. The release notes scope the print-only sentence to fix-21. Co-Authored-By: Claude <noreply@anthropic.com> Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
index.md, release-notes.md, and release.md open with a MyST front matter block, which MyST only reads from the first line. hawkeye only detects a header at the top of the file, so it flagged the headers that sit directly after the front matter. Co-Authored-By: Claude <noreply@anthropic.com> Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
juhovainio
left a comment
There was a problem hiding this comment.
Docs-only change, and I didn't find anything blocking. CI is green (lychee, Sphinx -W, hawkeye, Read the Docs, e2e, Windows); the self-hosted GPU lanes are skipping. This reflects a read of the diff, not a runtime check.
Things worth closing before or at merge (inline comments below):
- Release timing: the docs publish "0.3.0" ahead of the release.
conf.pyhard-codes it,release.mdsays "To be announced", the notes are based on rc.1, and the table links to adocs-0.3.0branch that doesn't exist yet. - Driver wording: the
install driveropener in the README should be checked against the code and againstwsl.md(does plain--yesapply the plan?). The release notes cover dnf/zypper while the README describes apt only. - Scope: this bundles copyediting, new release-notes pages, nav/landing restructure, the version bump and hawkeye excludes, and depends on the separate accuracy PR landing right after. It's a lot for one review, though the diff matches the PR description.
- Nits: the
wsl.mdheading rename changes its anchor for external deep links; a few acronyms (VRAM, OOM) are expanded after first use; some new README lines are unwrapped.
Remaining unchecked test-plan items (final-tag refresh, GIF render, branding and engineering sign-off) are the author's to close.
| # -- Project information --------------------------------------------------- | ||
|
|
||
| project = "ROCm CLI" | ||
| version = "0.3.0" |
There was a problem hiding this comment.
The version is hard-coded to 0.3.0 while 0.2.1 is the latest shipped release, so the header and nav title on the latest docs will read 0.3.0 before that release exists. Please sequence the merge with the release, or confirm this is intended.
There was a problem hiding this comment.
Intended: this PR targets release/v0.3.0, so the 0.3.0 version is meant to ship with that release. The test plan item to refresh from the final tag stays open until then.
|
|
||
| | Version | Release date | | ||
| | ------- | ------------ | | ||
| | [0.3.0](https://rocm.docs.amd.com/projects/rocm-cli/en/docs-0.3.0/index.html) | To be announced | |
There was a problem hiding this comment.
This links to the docs-0.3.0 branch, which doesn't exist yet. lychee runs offline, so CI can't catch a dead link here. Please confirm the URL pattern and branch before merge.
There was a problem hiding this comment.
Same as above. docs-0.3.0 appears when the branch is published, and the docs-0.2.1 row confirms the URL pattern. I'll check it after publishing.
|
|
||
| `install driver` installs the AMD kernel driver on Linux, using DKMS or a native | ||
| package. On bare metal it defaults to the floating `latest` amdgpu release; | ||
| `install driver` previews or installs Linux AMD driver support. Without `--dkms` |
There was a problem hiding this comment.
Could you confirm this against the driver-plan code? wsl.md says plain rocm install driver --yes runs the plan, while this reads as if --yes only applies together with --dkms. Also, the release notes mention dnf/zypper support while this section describes apt only.
There was a problem hiding this comment.
Checked against the code and fixed in 7c1b1e7. On WSL2 the plan is ROCDXG and --yes applies it without --dkms, so wsl.md is right and the README now says so. The README also describes the dnf (RHEL, Oracle Linux, Rocky Linux) and zypper (SLES) plans.
| ╚═╝ ╚═╝ ╚═════╝ ╚═════╝╚═╝ ╚═╝ ╚═════╝╚══════╝╚═╝ | ||
|
|
||
| Local AI on AMD GPUs — one binary, zero setup | ||
| Local AI on AMD GPUs — one install, zero setup |
There was a problem hiding this comment.
The tagline change ("one binary" -> "one install") needs branding sign-off, which is still unchecked in the test plan.
There was a problem hiding this comment.
That test-plan item was my own placeholder, so I have removed it. install.sh and install.ps1 still say "one binary, zero setup", which I am noting in case someone wants to align them.
| ".tmp-*/**", | ||
| "third_party/**", | ||
| "skills/rocm-doctor/**", | ||
| "docs/rocm-docs/index.md", |
There was a problem hiding this comment.
Excluding whole files means a future license-header regression in these pages won't be caught by the header check. Fine for now, just noting the trade-off.
r0x0r
left a comment
There was a problem hiding this comment.
I went through this closely, and I want to lead with the honest part: the prose is a real improvement. I compared the dense README and vLLM paragraphs clause by clause against the rewrites and could not find a single fact dropped or invented — the restructuring is faithful, not lossy. I also checked every {include} anchor in docs/rocm-docs/ against the rewritten README and all six still resolve. The "single prebuilt binary" to "rocm and rocmd" correction is right (xtask/src/package.rs:131 copies exactly those two), the new landing-page card text is traceable to the pages it links (the vLLM card matches the opening of docs/vllm.md), and CI is green on this head.
So this is close. What blocks it is narrow — roughly a one-line fix.
The new install driver opener is false on WSL2, and that opener is the one thing the PR title says it fixes. Details are inline; the short version is that the code takes the WSL branch and returns a mutating plan before --dkms is ever read, and docs/wsl.md — edited in this same PR — says the opposite, correctly.
The second blocker is the PR description rather than the diff: CONTRIBUTING.md picks up +34/-10 across two dedicated commits and is not mentioned at all. I left that inline too, anchored to a CONTRIBUTING.md hunk since a description has no diff line to hang a comment off.
There is also a non-blocking note about two hardcoded URLs that 404 today with nothing checking them.
Two things worth having on the record rather than changing:
licenserc.tomlnow excludes three docs pages by path. The headers are still in those files, but nothing enforces them there from here on — worth being a deliberate decision rather than a side effect.release.mdships "To be announced" for 0.3.0. Your own unchecked box already tracks it, so I am not asking for a change — just confirm the release-time follow-up has an owner.
|
|
||
| `install driver` installs the AMD kernel driver on Linux, using DKMS or a native | ||
| package. On bare metal it defaults to the floating `latest` amdgpu release; | ||
| `install driver` previews or installs Linux AMD driver support. Without `--dkms` |
There was a problem hiding this comment.
Blocking. This sentence is unqualified, but it is not true on WSL2 — and install driver is the opener the PR title is fixing.
In apps/rocm/src/driver_plan.rs, build_driver_install_plan_at takes the WSL branch and returns wsl_rocdxg_driver_plan(escalation) before dkms is ever read, and that plan carries mutating: true (driver_plan/rocdxg.rs:277). So on WSL2 there is no "without --dkms nothing happens" — a plain rocm install driver --yes installs ROCDXG. The paragraph has a Windows carve-out right here but no WSL one.
docs/wsl.md:40-48, edited in this same PR, states exactly that and is correct, so the two pages now contradict each other.
Suggested shape: qualify the non-mutating claim to bare metal — mirroring the --help text at cli.rs:782, "Preview or install Linux AMD driver support" — and add a sentence covering WSL2.
There was a problem hiding this comment.
Fixed in 7c1b1e7. The opener now scopes the non-mutating plan and the --dkms requirement to bare metal, and says that on WSL2 there is no DKMS flow: the command plans the ROCDXG install and --yes applies it without --dkms, with a link to the WSL page. I checked it against build_driver_install_plan_at and rocdxg.rs:277.
| `prek` runs the same checks locally that CI enforces: `cargo fmt`, `clippy`, `cargo test`, `ruff` (Python), `shellcheck` (shell), PowerShell syntax, license headers (`hawkeye`), markdown links (`lychee`), and the generated manifests (`MANIFEST.md`, `THIRD_PARTY_NOTICES.txt`). | ||
|
|
||
| The pre-push hooks (`cargo clippy`, the e2e harness clippy and `cargo test`) check only the crates your push affects: the crates whose files changed plus every crate that depends on them, selected the way CI's `cargo xtask affected` does. They check the whole workspace when a change cannot be confined to specific crates (the lockfile, the root manifest, the toolchain file, workflow files, a doc an xtask test reads, or a path no crate owns), and when the affected crates cannot be determined at all, for example because the remote branch tip is not fetched locally or the pushed range lists no changed files. They also add the crates whose tests read files from outside their own crate: any Rust file adds `xtask`, whose contract tests scan every source file, and `README.md`, `docs/` and `skills/` add `rocm` and `xtask`. Each step prints the scope it chose and why. A push that changes no Rust file runs none of the cargo steps, whatever else it changes (a lockfile-only push included). A green hook is an early warning and not the gate: on pull requests CI lints the whole workspace, but its Linux test job runs only the affected crates, and the whole workspace's tests run in the Windows debug job, in the merge queue and on `main`. To see what a push would check without compiling anything, run `cargo xtask pre-push-check test --dry-run --base origin/main` (the steps are `clippy`, `clippy-e2e` and `test`). | ||
| The pre-push hooks (`cargo clippy`, the e2e harness clippy, and `cargo test`) check only the crates your push affects: the crates whose files changed plus every crate that depends on them, selected the way CI's `cargo xtask affected` does. Each step prints the scope it chose and why. |
There was a problem hiding this comment.
Blocking, and this one is about the PR description rather than this line. I am anchoring here because a description has no diff line to comment on, and this hunk is part of the CONTRIBUTING.md work in question.
CONTRIBUTING.md takes +34/-10 across two dedicated commits: the pre-push-hook paragraph split into bullets (this hunk), bare paths converted to absolute GitHub links, new lead-ins before the tables and lists, text fences on the commit and branch examples. None of it appears in the description — no bullet mentions the file at all. "Docs only; no code changed" also glosses over the new Python log-filter handler in docs/rocm-docs/conf.py.
Unmentioned work is unreviewed work. Please add a bullet for CONTRIBUTING.md and narrow that final line.
There was a problem hiding this comment.
Added. The description now has a CONTRIBUTING.md bullet covering the pre-push paragraph split, lead-ins, fence language tags, absolute links, and wording changes, plus a bullet for the html-page-context hook in conf.py. The last line now reads "Docs and Sphinx config only (conf.py, licenserc.toml); no product code changed."
|
|
||
| | Version | Release date | | ||
| | ------- | ------------ | | ||
| | [0.3.0](https://rocm.docs.amd.com/projects/rocm-cli/en/docs-0.3.0/index.html) | To be announced | |
There was a problem hiding this comment.
Non-blocking. Two URLs added in this PR 404 today and nothing will catch it: this one (docs-0.3.0, docs/rocm-docs/release.md:19) and the version_list_link at docs/rocm-docs/conf.py:61, which points at .../en/latest/release.html.
I probed all three: the docs-0.2.1 row on the next line returns 200, which confirms the pattern is right, while docs-0.3.0 and latest/release.html both 404. lychee runs offline-only and skips docs/rocm-docs/ anyway, so there is no guard here.
Both resolve only once the docs are published, and the Version List link additionally needs a merge to main. Fine to merge as-is — just flagging that they are live-broken in the interim and unchecked.
There was a problem hiding this comment.
Agreed, and thanks for probing the pattern. Both 404s resolve once the docs are published (docs-0.3.0 after the branch is cut, latest/release.html after the merge to main). I'll check both after publishing.
…cronyms The install driver opener now says the DKMS flow applies to bare metal and that on WSL2 the command plans the ROCDXG install, where --yes applies it without --dkms. The driver section also describes the dnf and zypper plans. Four unwrapped README lines are rewrapped, and VRAM and OOM are expanded at first use in docs/vllm.md. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
The README link to docs/wsl.md sent site readers to GitHub. Split the commands page include around the opener and author a copy of it with the link retargeted to install/wsl.md, as install/wsl.md already does for its own retargeted sentence. Signed-off-by: pmoutsia_amdeng <peter.moutsias@amd.com>
Summary
README.md,docs/vllm.md, anddocs/wsl.md(single-sourced into the site): em dashes, "rather than", "e.g.", "vs.", British spellings, and the unwrapped paragraphs in the new "Howrocm servefinds the GPU on WSL" section. No new facts.install driveropener inREADME.md: it now says the command previews by default, that on bare metal without--dkmsit prints a non-mutating plan, and that--dkmsplans the DKMS install and--yesapplies it. On WSL2 there is no DKMS flow: the command plans the ROCDXG install and--yesapplies it without--dkms, with a link to the WSL page. The old "or a native package" wording is dropped because it could not be verified.amdgpu-dkms. Checked againstapps/rocm/src/driver_plan.rs. Why: the section only described apt-based distros.console.gifwas otherwise left unlinked from the site after the demos page was removed.####subheadings in Model serving (ports and hosts, other serve options, reuse, background and waiting, model form, GPU) and turn five dense paragraphs (--wait,--gpu, the reuse conditions, the ComfyUI reinstall behaviors, the--engine-envsecret rules) into lead-ins with bullets, and split the engine-options and lenient-matching paragraphs. Only sentence boundaries changed.index.rsttoindex.md. Replace the three category cards with four task cards that link to the same pages (WSL2 is a second link on the Install card), and show the CLI demo GIF after the intro, as the repository front page does. Card text comes from what those pages already say.conf.pyso the header and left nav title read ROCm CLI 0.3.0. The ROCm Core SDK link points at latest, as ondocs/0.2.1. The 0.3.0 release date is listed as "To be announced".rocmandrocmd) instead of "a single prebuilt binary" in the README and on the landing page (intro and Install card). Why: the accuracy PR corrects the install page to say prebuilt binaries, whichCargo.toml,install.sh, andrelease.ymlconfirm, so the README and landing page contradicted it. Merge this PR first and the accuracy PR immediately after, so the install page is not left saying "single" on its own.rocmandrocmd), so "one binary" no longer matches the docs.fix-21-msvc-redist-outdated) instead of all ofrocm fix. Why: only fix-21 is print-only (seeskills/rocm-doctor/reference.md), so the old wording read as a claim about every fix.docs/rocm-docs/index.md,release-notes.md, andrelease.mdfrom the hawkeye license header check inlicenserc.toml. These pages open with MyST front matter, which MyST only reads from the first line, and hawkeye only detects a header at the top of a file, so it flagged the headers that sit directly after the front matter. The headers are still in the files.CONTRIBUTING.md(two commits, about +34/-10): split the pre-push hook paragraph into short paragraphs and a bullet list; add lead-ins before the component table and the test, contribution, and review lists; addtextandbashlanguage tags to code fences; turn baredocs/paths into absolute GitHub links; expand IPC at first use; and replace "rather than" and "since" with plainer wording. No rules change.html-page-contexthook todocs/rocm-docs/conf.pythat points the ROCm Core SDK header link atlatest. Why: ondocs-X.Y.Zbranches rocm-docs-core resolves every sister project to that version, and the Core SDK has nodocs-0.3.0build.install driveropener and author a copy of that paragraph with the WSL link retargeted to the site's WSL page, asinstall/wsl.mdalready does for its own retargeted sentence. The README itself still links todocs/wsl.mdon GitHub. No other include anchors or hand-copied sentences changed.docs/vllm.md, and the file paths indocs/wsl.md(SME call). No new site pages.Test plan
sphinx-build -Wondocs/rocm-docspasses, including the converted landing page and nav.install driverwording checked againstapps/rocm/src/driver_plan.rsandapps/rocm/src/cli.rs.LD_LIBRARY_PATHbehavior indocs/wsl.mdmatches ROCMAI-870, ROCMAI-873: accept LD_LIBRARY_PATH for WSL ROCDXG readiness and give rocm serve WSL advice #915.wsl_librocdxgline stayingfalsefor a library found only throughLD_LIBRARY_PATH, and exit 1 outside a terminal.amdgpu-installis removed and dnf/zypper now use the repo-native flow (Remove amdgpu-install as an install method #181); the README driver section now describes both, checked againstdriver_plan.rs.install driveropener wordinggetting-started.htmlshows the CLI GIF after First run and the Console GIF after Interactive interfaces;demos.htmlis gone.rocm fixprint-only wording checked against the fix-21 entry inskills/rocm-doctor/reference.md.rocmandrocmdboth ship in the bundle, checked againstinstall.shandrelease.yml.hawkeye check(v7.0.0, as in CI) passes locally with the exclusions.release.mdand refresh the notes from the final tag (they are based on rc.1).docs-X.Y.Zlinks (inferred from the theme's ROCm CLI link).mediabranch, and that the landing page cards are not pushed too far down by the GIFDocs and Sphinx config only (
conf.py,licenserc.toml); no product code changed.