docs(rock4d/npu): add Kiln tutorial — local LLM + vision on a mainline kernel#1926
Open
gahingwoo wants to merge 1 commit into
Open
docs(rock4d/npu): add Kiln tutorial — local LLM + vision on a mainline kernel#1926gahingwoo wants to merge 1 commit into
gahingwoo wants to merge 1 commit into
Conversation
There was a problem hiding this comment.
Pull request overview
Adds a new, self-contained Kiln getting-started tutorial to the ROCK 4D NPU docs (ZH + EN), describing how to install Kiln’s mainline-kernel stack and use it for local LLM chat, vision inference, and an OpenAI-compatible API.
Changes:
- Add an English tutorial page covering install, verification (
kiln-doctor), vision (kiln-vision), chat (kiln-chat), serving (kiln-serve), conversion (kiln-convert), and troubleshooting. - Add a Chinese counterpart with equivalent flow and references.
- Include expected command outputs and operational tips (reboots, server host binding, version pinning).
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.
| File | Description |
|---|---|
| docs/rock4/rock4d/application-dev/npu/kiln.md | New Chinese Kiln tutorial page for ROCK 4D NPU on a mainline kernel. |
| i18n/en/docusaurus-plugin-content-docs/current/rock4/rock4d/application-dev/npu/kiln.md | New English Kiln tutorial page mirroring the Chinese content and flow. |
Comment on lines
+57
to
+66
| ## 1. Install | ||
|
|
||
| Run this on the ROCK 4D: | ||
|
|
||
| ```bash | ||
| curl -fsSL https://raw.githubusercontent.com/gahingwoo/kiln/main/scripts/kiln-install.sh | bash | ||
| ``` | ||
|
|
||
| It is hands-off: it pre-downloads what it needs, installs the Kiln mainline kernel, then | ||
| **reboots itself twice (~10–15 minutes total)** to finish setup and land in a ready system. |
Comment on lines
+42
to
+60
| ## Prerequisites | ||
|
|
||
| - 一块 **ROCK 4D（RK3576）**，运行 **Armbian**。 | ||
| - 可用网络（首次安装需要联网下载内核与运行时；**以太网最稳**）。 | ||
| - 模型自备（Kiln 不附带任何模型）： | ||
| - **视觉**：无需自备，下面用 `kiln-convert` 在板上现转一个 MobileNet 分类器。 | ||
| - **大模型**：准备一个与 `librkllmrt` **1.2.0** 匹配的 `*-rk3576-w4a16.rkllm`（例如 Qwen2.5-1.5B | ||
| 或 Llama-3.2-1B），稍后放进 `/opt/models`。 | ||
|
|
||
| ## 1. 安装 | ||
|
|
||
| 在 ROCK 4D 上执行这条命令： | ||
|
|
||
| ```bash | ||
| curl -fsSL https://raw.githubusercontent.com/gahingwoo/kiln/main/scripts/kiln-install.sh | bash | ||
| ``` | ||
|
|
||
| 它是**免手动**的：预下载所需内容、装上 Kiln 主线内核，然后**自己重启两次（共约 10–15 分钟）**完成 | ||
| 安装并进入就绪系统。 |
Comment on lines
+74
to
+77
| :::tip | ||
| To drive the reboots yourself, use manual mode: `KILN_MANUAL=1 bash kiln-install.sh` (it | ||
| tells you when to reboot and re-run). | ||
| ::: |
Comment on lines
+67
to
+69
| :::tip | ||
| 想自己掌控重启，用手动模式：`KILN_MANUAL=1 bash kiln-install.sh`（它会告诉你何时重启、何时再跑一遍）。 | ||
| ::: |
gahingwoo
force-pushed
the
tutorial/rock4d-npu-kiln
branch
from
4c96c1c to
d47ff62
Compare
A self-contained getting-started page for running a local LLM chat, image recognition, and an OpenAI-compatible API on the ROCK 4D NPU with Kiln, a community project that runs the vendor RKLLM/RKNN stack on a mainline linux-7.1.3 kernel. Covers install -> verify -> vision -> LLM chat -> Open WebUI -> model conversion -> troubleshooting, each with expected output. Armbian one-command install; clearly attributed as a third-party project. Adds zh (docs/) + en (i18n/en/) pages. Passes scripts/agent-doc-lint.sh and scripts/agent-doc-drift-guard.sh.
Contributor
Author
|
Thanks for the review — addressed both points:
Also tidied the reference links. |
gahingwoo
force-pushed
the
tutorial/rock4d-npu-kiln
branch
from
d47ff62 to
a39104d
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a self-contained getting-started tutorial to the ROCK 4D NPU docs, in both languages:
docs/rock4/rock4d/application-dev/npu/kiln.md(zh)i18n/en/docusaurus-plugin-content-docs/current/rock4/rock4d/application-dev/npu/kiln.md(en)It walks through running a private, offline local LLM chat + image recognition + an
OpenAI-compatible API on the ROCK 4D NPU using Kiln,
an open-source community project that runs the vendor RKLLM/RKNN NPU stack on a mainline
linux-7.1.3kernel (rather than the vendor 6.1 BSP).Scope
kiln-doctor) → vision (kiln-vision) →LLM chat (
kiln-chat) → Open WebUI / OpenAI API (kiln-serve) → model conversion(
kiln-convert) → troubleshooting, each with expected output.noted as validated on the ROCK 4D (RK3576).
Checks
scripts/agent-doc-lint.sh— passed (front matter + H1 + language-labeled code fences,no placeholder tokens).
scripts/agent-doc-drift-guard.sh— passed (zh + en added together, no path drift).AGENT-DOC-SPEC.md(Prerequisites / Steps / ExpectedOutput / Validation / Troubleshooting).
Happy to adjust the wording, placement (
sidebar_position), or framing per your review.