Skip to content

[feature] 迁移 onnxruntime Quick Start 并接入看护 - #169

Open
licy666 wants to merge 4 commits into
Ascend:mainfrom
licy666:feat/onnxruntime-quick-start-a2
Open

licy666 wants to merge 4 commits into
Ascend:mainfrom
licy666:feat/onnxruntime-quick-start-a2

Conversation

@licy666

@licy666 licy666 commented Sep 10, 2026

Copy link
Copy Markdown
Contributor

Summary

  • 把 cosdt-ci-test/workflows 里的 onnxruntime 昇腾入门文档迁到本仓:用户文档、测试子类、薄触发器、首页卡片在同一个 PR。
  • 文档按当前正式 GitHub Release 源码编译 onnxruntime-cann 再跑最小加法推理,不钉死 PyPI 轮子;否则绿灯对不上 microsoft/onnxruntime 的最新 tag。上次因此关掉过 [docs] onnxruntime 快速上手改为链到官方 CANN EP 文档 #153。
  • 面向香港 A2 runner:直连官方 GitHub、不传 container_options / --volume,跨 run 复用只放隐藏 #test-setup 的 /root/.cache/huggingface/onnxruntime/。推理块预期含 CANNExecutionProvider,并关掉 CPU 回退。

Test plan

  • 本机 actionlint、py_compile、解析块数 15 / 预期 8(防空跑)
  • 可见正文泄漏检查:无 /root/.cache、cosdt-ci-test、代理、ModelScope、:::{note}
  • coder 上按文档字面:克隆并编译 v1.29.1 约 21 分钟,安装与造模型通过,get_available_providers() 含 CANNExecutionProvider
  • coder 缺 /usr/local/Ascend/driver,aclInit 失败,推理未在该环境跑通;合入后在本仓 A2 runner 上 workflow_dispatch 再验冷/热缓存
  • 热缓存:隐藏 restore 命中后,可见编译块应跳过;把缓存改成脏数据后必须删掉再编

@ascend-robot

Copy link
Copy Markdown

CLA Signature Pass

licy666, thanks for your pull request. All authors of the commits have signed the CLA. 👍

github-actions Bot pushed a commit that referenced this pull request Sep 10, 2026
@github-actions

github-actions Bot commented Sep 10, 2026 •

Copy link
Copy Markdown
Contributor

📖 文档预览:https://ascend.github.io/docs/pr-preview/pr-169/sources/onnxruntime/

也可打开整站首页。

构建自 e0af114aad58726611240143579492221c4181b1。
每次推送后自动更新,PR 关闭时自动清理。

@ascend-robot

Copy link
Copy Markdown

CLA Signature Pass

licy666, thanks for your pull request. All authors of the commits have signed the CLA. 👍

从当前正式 Release 源码编译 onnxruntime-cann,避免钉死 PyPI 轮子后绿灯对不上上游 tag。
@ascend-robot

Copy link
Copy Markdown

CLA Signature Pass

licy666, thanks for your pull request. All authors of the commits have signed the CLA. 👍

github-actions Bot pushed a commit that referenced this pull request Sep 21, 2026
@Devin-DC-Huang

Copy link
Copy Markdown
Contributor
image

@Devin-DC-Huang

Copy link
Copy Markdown
Contributor
image 这里结果也要刷新,不仅仅是输出文件路径对比

@Devin-DC-Huang

Copy link
Copy Markdown
Contributor
image

读者不再先把脚本存成文件。探测步骤的预期留在看护里,页面展示设备样例、模型结构和加法结果。
@ascend-robot

Copy link
Copy Markdown

CLA Signature Pass

licy666, thanks for your pull request. All authors of the commits have signed the CLA. 👍

@ascend-robot

Copy link
Copy Markdown

CLA Signature Pass

licy666, thanks for your pull request. All authors of the commits have signed the CLA. 👍

github-actions Bot pushed a commit that referenced this pull request Sep 24, 2026
当前 onnx 默认写出 IR 14,onnxruntime 1.30 最多读到 13,推理会话因此打不开模型。
@ascend-robot

Copy link
Copy Markdown

CLA Signature Pass

licy666, thanks for your pull request. All authors of the commits have signed the CLA. 👍

github-actions Bot pushed a commit that referenced this pull request Sep 24, 2026

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants