AKG(Auto Kernel Generator) is an optimizer for operators in Deep Learning Networks, which provides the ability to automatically fuse ops with specific patterns.

zhengqishui1!2171 merge pto into br_agents

9f81757a创建于 1 天前3,011次提交

文件	最后提交记录	最后更新时间
.gitcode	update template	1 个月前
.github	init	5 年前
.jenkins	Resolve static code check	19 天前
akg-mlir	fix bug: loop fusion	3 个月前
akg_agents	add PyPTO anti-cheat to CodeChecker + document kernel output rules	1 天前
docs	update akg master	6 个月前
mfusion	Change pybind11 version	19 天前
.clang-format	update akg master	6 个月前
.gitattributes	update akg master	6 个月前
.gitignore	gitignore: stop ignoring workspace_autoresearch's .claude/ subtree The generic `.claude/` rule (two entries in this .gitignore) was meant for editor-local Claude Code config, the kind that lives under user home and shouldn't be in the repo. workspace_autoresearch ships project-level Claude Code definitions under its own `.claude/`: - settings.json points AKG_AGENTS_AR_SKILLS_ROOT at the in-repo skills tree and carries the AR-script permissions allowlist - commands/autoresearch.md is the /autoresearch slash command spec - agents/ar-diagnosis.md (added in follow-up commit) is the DIAGNOSE subagent prompt These are shared definitions, not personal config, so they belong in git. Negation rule sits after both generic .claude/ entries because gitignore is last-match-wins; the companion `/**` pattern is needed because a parent-dir negation alone does not always re-include its children.	5 天前
.gitmodules	refactor: unify benchmark downloads and remove submodules	1 个月前
CODEOWNERS	Update the codeowners for mfusion	20 天前
CONTRIBUTING.md	init	5 年前
LICENSE	init	5 年前
NOTICE	update akg master	6 个月前
OWNERS	update akg master	6 个月前
README.md	update README	1 个月前
README_en.md	update README	1 个月前
Third_Party_Open_Source_Software_Notice	remove ncurses dependency	1 个月前
version.txt	update akg master	6 个月前

Auto Kernel Generator (AKG)

项目简介

AKG(Auto Kernel Generator)是一个深度学习编译器，用以对深度神经网络中的算子进行优化，并提供特定模式下的算子自动融合功能。从 2.4 版本之后，我们对 AKG 进行了全新升级，当前的 AKG 项目中，包含了 AKG-AGENT、AKG-MLIR 2个子项目，支持 CPU、NVIDIA、V100/A100、Atlas 800T A2/A3、300I DUO 等多硬件后端的算子自动生成。

AKG-AGENT： AKG-AGENT是我们最新孵化的一款大模型驱动的算子生成工具，提供了基于 Multi-agent 的算子自动生成框架、AscendKernelBench 评测平台、以及算子知识库。当前支持通过 LLM 的方式，生成 Triton-Ascend（Atalas A2/A3）以及 SWFT（300I DUO）的 DSL，并完成算子 kernel 编译。
AKG-MLIR： 基于 MLIR 开源项目演进的深度学习编译器，提供了 CPU/GPU/Ascend 上完整的算子编译 Pipeline。当前包含了 MindSpore Dialect 图编译方言，并对 Linalg、Affine、GPU 等方言进行了扩展，增强了循环融合调度能力。此外，AKG-MLIR 对接了 AscendNPU IR，支持昇腾后端融合算子生成。