mindspore-ai / mindspore-ai/hyper-parallel

新增分布式算子开发技能(dist-op-analysis / dist-op-dev)及开发流程简化

Open
#712 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
53
Forks
63
Avg merge
23h 45m
Merged PRs (30d)
63

Description

背景

HyperParallel 的分布式算子开发流程缺乏标准化工具支撑。开发者需要手动查阅大量分散的文档(YAML 注册规范、preprocess/infer_layout/get_expand_impl 编码约束、UT/ST 测试模式),新算子接入效率低且容易遗漏检查项。同时,原有的 dist-op-dev 技能包含大量冗余的 workflow 文档(6 个独立 workflow MD 文件),与已有规则文件内容重叠,维护成本高。

目标

重构并精简分布式算子开发技能,提供清晰的端到端流程引导:

dist-op-analysis(算子分析)
  • 接受用户提供的 PyTorch/MindSpore 接口信息,输出标准化的算子分析报告和实现计划
  • 提供 plan-template.md 模板,统一输出的计划文件格式,包含:接口定义、分布式策略、分片场景表格、实现方案
  • 需人工确认后才进入实现阶段/dist-op-analysis → 输出 plan → STOP → 人工确认 → /dist-op-dev
dist-op-dev(算子开发)
  • 读取已确认的 plan 文件,自动执行实现 → 测试 → 修复循环,直到所有可执行测试通过
  • 简化了技能文件结构:
    • 移除 6 个独立的 workflow MD 文件(01 分析 → 06 提交流程),所有约束合并到 2 个规则文件
    • 移除与 ../rules/ 重复的 3 个参考文件
    • 新增 plan-template.md 模板统一分析报告与实现计划的格式
  • 验证报告保存到 reports/ 目录(.gitignore 排除,本地留存)

相关 PR

#695

schema_version: 1
source: gitcode
gitcode_repo: mindspore/hyper-parallel
gitcode_issue: 167
source_url: https://gitcode.com/mindspore/hyper-parallel/issues/167

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reviewing the existing dist-op-dev skill, its six workflow Markdown files, the ../rules/ references, and the related PR #695. Compare the requested dist-op-analysis flow, dist-op-dev simplification, plan-template.md, and reports/ handling against the current structure. Done means the consolidated skills support the documented analysis-confirmation-development-test cycle without duplicated workflow material.

Written by the indexing model from the issue text.

Assessment

Tech stack
pytorch
Domain
documentation, tooling
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.