mindspore-ai/mindspore-lite

open-source-model-migration

把开源算法模型适配到 MindSpore Lite 部署管线：按网络结构拆分导出 ONNX、ONNX Runtime 推理验证、ONNX→MindIR 转换、MindSpore Lite 推理实现，并交付文档与常见问题。用户想把某个开源模型迁移到 MSLite 部署时调用。

Source text: Chinese

precision-troubleshooting

MindSpore Lite MindIR模型精度问题定位技能。覆盖从CANN Profiling对比分析定位可疑算子、配置算子Dump、到最终对比Dump数据确定精度差异根因算子的全流程。

Source text: Chinese

tensor-parallel-deployment

开源 LLM 模型的多卡张量并行部署：Megatron 风格权重分片导出 → ONNX→MindIR 转换（离线 acl / GE online 双路径）→ 多进程 HCCL 推理。覆盖 1p/2p/4p 三种模式，含 Custom 算子（AllReduce/Scatter/RMSNorm 等）开发与通信机制、300I Duo 与 800I A2 差异、已知限制与修复方案。用户想把模型部署到多卡/多芯 Ascend 推理时调用。

Source text: Chinese

performance-optimization

MindSpore Lite（Ascend）模型性能优化总攻略。做基线/profiling、融合算子改写、推理免拷贝、PTQ int8 量化、精度对齐与归档时调用。本文为总览与索引，细化策略见 references/。

Source text: Chinese

onnx-model-conversion-and-deployment

MindSpore Lite云侧推理 Ascend 后端离线转换（ONNX → MindIR）与推理部署全流程。覆盖固定 shape、动态分档、纯动态 shape 的转换策略，以及 MindIR 推理验证与部署注意事项。

Source text: Chinese

lite-build

Build configuration, CMake options, cross-compilation and packaging. Use when building MindSpore Lite, configuring CMake, cross-compiling for ARM/iOS/MCU, packaging release archives, or troubleshooting build errors.

lite-converter

Model conversion pipeline, parser development, optimization passes and quantization. Use when converting models to .ms, writing parser code, implementing optimizer passes, or configuring quantization.

lite-debug-test

Debugging, unit testing, benchmarking and performance analysis. Use when running gtest, benchmark tools, profiling latency or accuracy, diagnosing operator precision issues, delegate fallback, or memory leaks.

lite-device-side-infer

Device-side inference with LiteRT, NNACL and hardware delegates. Use for mobile/IoT inference, Android/iOS integration, NPU/GPU/CoreML delegates, Micro codegen for MCU, on-device training, or C/C++/Java/Python API usage with .ms models.

lite-kernel-dev

Operator and kernel development, NNACL, delegates, custom kernel registration. Use when adding operators, implementing NNACL kernels, writing delegate adapters (NPU/CoreML/Ascend), registering custom kernels, or modifying operator schema.

clean-code-check

改完或 review MindSpore Lite 的 C++/Python/Shell/CMake 代码时用。先跑 pre-push（8 个 lint 全自动），再按 30 秒快速清单逐项过——三类硬指标（CCN≤19 / NLOC≤50 / 入参≤5）+ 指针/边界/逻辑/异常/接口契约等 lint 抓不到的语义问题 + 涉及外部数据时的安全编码。深度参考 clean-code-guidelines.md。

Source text: Chinese