TorchAir 支持用户基于PyTorch框架和torch_npu插件在昇腾NPU上使用图模式进行推理。

文件	最后提交记录	最后更新时间
.claude	code review skill Co-authored-by: fu-hao-huawei<fuhao16@huawei.com> # message auto-generated for no-merge-commit merge: !3246 merge review_skill into master code review skill Created-by: fu-hao-huawei Commit-by: fu-hao-huawei Merged-by: ascend-robot Description: code review skill See merge request: Ascend/torchair!3246	24 天前
.gitcode	rename .gitee to .gitcode Co-authored-by: stevenaw0<huangguijun@huawei.com> # message auto-generated for no-merge-commit merge: !3111 merge master into master rename .gitee to .gitcode Created-by: stevenaw0 Commit-by: stevenaw0 Merged-by: ascend-robot Description: rename .gitee to .gitcode See merge request: Ascend/torchair!3111	1 个月前
.opencode	docs: 修复claude启动时远端skill下载失败问题 Co-authored-by: stevenaw0<huangguijun@huawei.com> # message auto-generated for no-merge-commit merge: !3109 merge plugin into master docs: 修复claude启动时远端skill下载失败问题 Created-by: stevenaw0 Commit-by: stevenaw0 Merged-by: ascend-robot Description: ## 变更内容 1. 新增 AGENTS.md：中文项目指引文档，包含项目概述、构建命令、测试方法、架构说明、核心 API 等 2. 新增 .claude/CLAUDE.md：指向 AGENTS.md 的引导文件 3. 下载远端skill的脚本增加可执行权限：修复claude启动时自动下载失败问题： 4. .claude/settings.json增加舱内曾删改查权限 ## Checklist - [x] 代码已通过本地测试 - [x] 提交信息符合规范 - [x] 无敏感信息泄露 See merge request: Ascend/torchair!3109	1 个月前
cmake	clean code Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !3171 merge clean_code_clang-format into master clean code Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: clean_code_clang-format c++ clean code See merge request: Ascend/torchair!3171	1 个月前
codegen	pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!2931	2 个月前
docs	(master)torch npu is renamed to TorchNPU Co-authored-by: lyz0328<liyanzhen@huawei.com> # message auto-generated for no-merge-commit merge: !3359 merge master into master (master)torch npu is renamed to TorchNPU Created-by: lyz0328 Commit-by: lyz0328 Merged-by: ascend-robot Description: (master)torch npu is renamed to TorchNPU See merge request: Ascend/torchair!3359	7 天前
examples	open ruff rule of pre-commit Co-authored-by: 冯振强<fengzhenqiang@huawei.com> # message auto-generated for no-merge-commit merge: !3371 merge pre_commit_test into master open ruff rule of pre-commit Created-by: feng_zhenqiang Commit-by: 冯振强 Merged-by: ascend-robot Description: open ruff rule of pre-commit See merge request: Ascend/torchair!3371	2 天前
experimental	add dim>=4 skip decompose Co-authored-by: yurongkun<yurongkun@huawei.com> # message auto-generated for no-merge-commit merge: !3405 merge add_4_axis_handle into master add dim>=4 skip decompose Created-by: yurongkun Commit-by: yurongkun Merged-by: ascend-robot Description: add softmax axis <= 3 can decompose handle add Merge branch 'add_softmax_opt' into add_4_axis_handle add dim>=4 skip decompose See merge request: Ascend/torchair!3405	18 小时前
feature	(master)torch npu is renamed to TorchNPU Co-authored-by: lyz0328<liyanzhen@huawei.com> # message auto-generated for no-merge-commit merge: !3359 merge master into master (master)torch npu is renamed to TorchNPU Created-by: lyz0328 Commit-by: lyz0328 Merged-by: ascend-robot Description: (master)torch npu is renamed to TorchNPU See merge request: Ascend/torchair!3359	7 天前
npugraph_ex	fix C0123 C0201 C0206 C0207 C0302 R1711 Co-authored-by: 冯振强<fengzhenqiang@huawei.com> # message auto-generated for no-merge-commit merge: !3401 merge pre_commit_test into master fix C0123 C0201 C0206 C0207 C0302 R1711 Created-by: feng_zhenqiang Commit-by: 冯振强 Merged-by: ascend-robot Description: \| 工具 \| 错误码 \| 处理数量 \| 处理方式 \| 验证结果 \| \| --- \| --- \| ---: \| --- \| --- \| \| Pylint \| C0123 \| 1 \| 将 `type(...) != ...` 改为 `not isinstance(...)` \| 单文件验证无 C0123 \| \| Pylint \| C0201 \| 53 \| 将 `x in dict.keys()` 改为 `x in dict` \| 定向验证无 C0201 \| \| Pylint \| C0206 \| 7 \| 将需要值的字典遍历改为 `dict.items()` \| 定向验证无 C0206 \| \| Pylint \| C0207 \| 2 \| 将 `split('.')[-1]` 改为 `rsplit('.', maxsplit=1)[-1]` \| 定向验证无 C0207 \| \| Pylint \| C0302 \| 10 \| 在触发文件顶部添加模块级 `too-many-lines` 屏蔽；未在配置中全局屏蔽 \| 定向验证无 C0302 \| \| Pylint \| R1711 \| 105 \| 删除函数末尾无效 `return` / `return None`，必要处改为 `pass` \| `--enable=R1711` 全量验证通过 \| See merge request: Ascend/torchair!3401	1 天前
pre-commit	opend ruff check Co-authored-by: 冯振强<fengzhenqiang@huawei.com> # message auto-generated for no-merge-commit merge: !3393 merge pre_commit_test into master opend ruff check Created-by: feng_zhenqiang Commit-by: 冯振强 Merged-by: ascend-robot Description: opend ruff check See merge request: Ascend/torchair!3393	2 天前
proto	support hif8 for graph Co-authored-by: huangfei<huangfei32@huawei.com> # message auto-generated for no-merge-commit merge: !2145 merge master into master [Feature] [Feature] support hif8 for graph Created-by: dddxhh Commit-by: huangfei Merged-by: ascend-robot Description: <!-- 请根据您的Pull Request类型，保留以下相应的一个模板部分，并删除其他无关部分 --> <!-- 请根据您的Pull Request类型，在Pull request标题中标记[Feature]\[Bugfix]\[Refactor] --> <!-------------------- 新功能模板 BEGIN --------------------> ### 【新功能提交】功能描述 support hif8 for graph 相关Issue <!-- 如有关联的Issue，请在此处填写，例如: [Fixes issue xxx](链接) --> Checklist - [x] 符合编程规范&&代码红线 - [x] 具备UT测试用例看护&&测试用例为有效用例（若无测试用例看护请说明原因） - [ ] 是否涉及对外接口变更 - [ ] 是否涉及官网文档修改 <!-------------------- 新功能模板 END --------------------> See merge request: Ascend/torchair!2145	7 个月前
python	A5将不再支持npu_moe_gating_top_k_softmax接口，引导用户切换至npu_moe_gating_top_k_softmax_v2 Co-authored-by: ZhouChuping<zhouchuping1@huawei.com> # message auto-generated for no-merge-commit merge: !3370 merge master into master A5将不再支持npu_moe_gating_top_k_softmax接口，引导用户切换至npu_moe_gating_top_k_softmax_v2 Created-by: ZhouChuping Commit-by: ZhouChuping Merged-by: ascend-robot Description: 关联issue https://gitcode.com/Ascend/op-plugin/issues/268 A5将不再支持npu_moe_gating_top_k_softmax接口，引导用户切换至npu_moe_gating_top_k_softmax_v2 将npu_moe_gating_top_k_softmax 接口的硬阻断（报错）改为软弃用（告警）功能验证ok，改为仅打印日志提示 ![image.png](https://raw.gitcode.com/user-images/assets/7964479/afa4e5a3-f286-4a5f-bf42-1296b11ad585/image.png 'image.png') See merge request: Ascend/torchair!3370	1 天前
tests	Revert "force eager add clone input and output" Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !3334 merge force_eager_back into master Revert "force eager add clone input and output" Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: Revert "fix bug capture host input" Revert "force eager no need getCheckpointState" Revert "force eager pool" Revert "force eager add clone input and output" See merge request: Ascend/torchair!3334	10 天前
third_party	delete quant_all_reduce & quant_reduce_scatter Co-authored-by: yifux<xiongyifu1@huawei.com> # message auto-generated for no-merge-commit merge: !3377 merge pr_del into master delete quant_all_reduce & quant_reduce_scatter Created-by: xiongyifu Commit-by: yifux Merged-by: ascend-robot Description: delete quant_all_reduce & quant_reduce_scatter See merge request: Ascend/torchair!3377	2 天前
tools	[feature]npugraph_ex support dead lock check Co-authored-by: chenweiheng<chenweiheng1@hisilicon.com> # message auto-generated for no-merge-commit merge: !2782 merge master into master [feature]npugraph_ex support dead lock check Created-by: chenweiheng Commit-by: chenweiheng Merged-by: ascend-robot Description: [feature]npugraph_ex support dead lock check See merge request: Ascend/torchair!2782	3 个月前
torchair	clean code Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !3171 merge clean_code_clang-format into master clean code Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: clean_code_clang-format c++ clean code See merge request: Ascend/torchair!3171	1 个月前
.clang-format	pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!2931	2 个月前
.gitignore	support cat lowering Co-authored-by: medivh<xuepeng4@huawei.com> # message auto-generated for no-merge-commit merge: !3139 merge master into master support cat lowering Created-by: medivh-x Commit-by: medivh Merged-by: ascend-robot Description: support cat lowering See merge request: Ascend/torchair!3139	1 个月前
.gitmodules	update submodiles Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !3375 merge update_submodel_url into master update submodiles Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: 社区要求三方库链接切换使用问题的gitcode网站 See merge request: Ascend/torchair!3375	3 天前
.pre-commit-config.yaml	opend ruff check Co-authored-by: 冯振强<fengzhenqiang@huawei.com> # message auto-generated for no-merge-commit merge: !3393 merge pre_commit_test into master opend ruff check Created-by: feng_zhenqiang Commit-by: 冯振强 Merged-by: ascend-robot Description: opend ruff check See merge request: Ascend/torchair!3393	2 天前
AGENTS.md	docs: 补充项目概述中 npugraph_ex ACL 图执行路径的描述 Co-authored-by: stevenaw0<huangguijun@huawei.com> # message auto-generated for no-merge-commit merge: !3123 merge add_npugraph_ex into master docs: 补充项目概述中 npugraph_ex ACL 图执行路径的描述 Created-by: stevenaw0 Commit-by: stevenaw0 Merged-by: ascend-robot Description: ## 变更说明更新 AGENTS.md 项目概述，补充 npugraph_ex 提供的 ACL Graph 捕获执行路径描述，使文档完整反映 TorchAir 的两条图编译后端（GE 全图编译 + ACL 捕获执行）。 ## 变更内容 - 修改 AGENTS.md 第 5 行项目概述，增加 ACL Graph 捕获执行路径说明 See merge request: Ascend/torchair!3123	1 个月前
CMakeLists.txt	support ascendc packaging to torch_npu/_inductor Co-authored-by: dingdairong<dingdairong@huawei.com> # message auto-generated for no-merge-commit merge: !3329 merge inductor_af_ascendc_packaging into master support ascendc packaging to torch_npu/_inductor Created-by: dingdairong Commit-by: dingdairong Merged-by: ascend-robot Description: support ascendc packaging to torch_npu/_inductor 只在编译torch_npu包的时候，打包inductor_npu-ext打包到torch_npu/_inductor。独立torchari whl包安装时，不做此动作。 See merge request: Ascend/torchair!3329	13 天前
CONTRIBUTING.md	(master)torch npu is renamed to TorchNPU Co-authored-by: lyz0328<liyanzhen@huawei.com> # message auto-generated for no-merge-commit merge: !3359 merge master into master (master)torch npu is renamed to TorchNPU Created-by: lyz0328 Commit-by: lyz0328 Merged-by: ascend-robot Description: (master)torch npu is renamed to TorchNPU See merge request: Ascend/torchair!3359	7 天前
LICENSE	pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!2931	2 个月前
OWNERS	pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!2931	2 个月前
README.en.md	fix docs bug Co-authored-by: hulida<hulida@huawei.com> Co-authored-by: fu-hao-huawei<fuhao16@huawei.com> Co-authored-by: chenweiheng<chenweiheng1@hisilicon.com> Co-authored-by: SunYaping<sunyaping4@huawei.com> Co-authored-by: Achilles_d<duanchengyuan@huawei.com> Co-authored-by: zhukkk<zhuke11@huawei.com> Co-authored-by: tangjie66<tangjie66@huawei.com> # message auto-generated for no-merge-commit merge: !2811 merge ass into master fix docs bug Created-by: Sunyp01 Commit-by: Sunyp01;Achilles_d;fu-hao-huawei;chenweiheng;hulida;tangjie66;zhukkk;SunYaping Merged-by: ascend-robot Description: fix doc bug See merge request: Ascend/torchair!2811	3 个月前
README.md	(master)torch npu is renamed to TorchNPU Co-authored-by: lyz0328<liyanzhen@huawei.com> # message auto-generated for no-merge-commit merge: !3359 merge master into master (master)torch npu is renamed to TorchNPU Created-by: lyz0328 Commit-by: lyz0328 Merged-by: ascend-robot Description: (master)torch npu is renamed to TorchNPU See merge request: Ascend/torchair!3359	7 天前
RELEASE.md	torch air framework supported now	2 年前
SECURITY_README.md	(master)torch npu is renamed to TorchNPU Co-authored-by: lyz0328<liyanzhen@huawei.com> # message auto-generated for no-merge-commit merge: !3359 merge master into master (master)torch npu is renamed to TorchNPU Created-by: lyz0328 Commit-by: lyz0328 Merged-by: ascend-robot Description: (master)torch npu is renamed to TorchNPU See merge request: Ascend/torchair!3359	7 天前
Third_Party_Open_Source_Software_Notice	clean code Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !3171 merge clean_code_clang-format into master clean code Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: clean_code_clang-format c++ clean code See merge request: Ascend/torchair!3171	1 个月前
build.sh	support all existed asc ops Co-authored-by: medivh<xuepeng4@huawei.com> # message auto-generated for no-merge-commit merge: !2592 merge master into master support all existed asc ops Created-by: medivh-x Commit-by: medivh Merged-by: ascend-robot Description: 按 pyascir.h REGISTERED_OPS / ascir_builtin_ops.cpp 把当前 ascir 已暴露的 pointwise / reduce 算子在 asc_ops.py 补齐 wrapper，覆盖 65 个 op： - unary：三角函数、对数 / 指数族、isnan/isfinite/relu/gelu/leaky_relu、 bitwise_not / logical_not、nop - unary 类型转换：floor_to_int / ceil_to_int / round_to_int / trunc_to_int - binary：minimum / atan2 / copysign / hypot / fmod / mod / remainder / floordiv / truncdiv / axpy / bitwise_or/xor / xor / logical_and/or/xor / lshift / rshift（+ 别名） - ternary：fma / select / clip_by_value 补齐后 NPUOverrides 中 ir.tanh / ir.cos / ir.atan2 等不再走 __getattr__ 回退到 _unsupported_op，能真实落到 ascir。 lowering 白名单同步放开：aten.log1p / aten.relu / aten.sgn / aten.sign。其中 sgn 在实数路径下 inductor 把它 decompose 成 ops.sign，所以两个都要放行。 _finetune_decompose 白名单从单 _to_copy 扩成额外保留 aten.silu / aten.sgn 两个 OpOverloadPacket 的全部 overload —— inductor 没给这两个 op 注册直接 lowering，必须靠 decomposition 拆成 NPUOverrides 支持的原子 op: - silu(x) = x / (1 + (-x).exp()) → neg/exp/add/div - sgn(x) → sign(x)（实数路径）否则 cpu 模式（不带 'decompose' / 'lowering' flag）下整个 op 进 lowering 找不到实现，会 fallback 回 aten。不含 Gather / Concat / IndexExpr —— 这几个 ascir 当前实现有问题，留给 _unsupported 回退或上层自定义 lowering。 See merge request: Ascend/torchair!2592	1 个月前
build_and_install.sh	optimize torchair and pta ci time Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2987 merge ci_times into master optimize torchair and pta ci time Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: optimize torchair and pta ci time See merge request: Ascend/torchair!2987	2 个月前
classify_rule.txt	add release/ut/st ci build and coverage generate	2 年前
classify_rule.yaml	pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!2931	2 个月前
configure	optimize torchair and pta ci time Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2987 merge ci_times into master optimize torchair and pta ci time Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: optimize torchair and pta ci time See merge request: Ascend/torchair!2987	2 个月前
configure.py	block W1514 Co-authored-by: 冯振强<fengzhenqiang@huawei.com> # message auto-generated for no-merge-commit merge: !3397 merge pre_commit_test into master block W1514 Created-by: feng_zhenqiang Commit-by: 冯振强 Merged-by: ascend-robot Description: block W1514 See merge request: Ascend/torchair!3397	2 天前
public_address_statement.md	fix docs bug Co-authored-by: hulida<hulida@huawei.com> Co-authored-by: fu-hao-huawei<fuhao16@huawei.com> Co-authored-by: chenweiheng<chenweiheng1@hisilicon.com> Co-authored-by: SunYaping<sunyaping4@huawei.com> Co-authored-by: Achilles_d<duanchengyuan@huawei.com> Co-authored-by: zhukkk<zhuke11@huawei.com> Co-authored-by: tangjie66<tangjie66@huawei.com> # message auto-generated for no-merge-commit merge: !2811 merge ass into master fix docs bug Created-by: Sunyp01 Commit-by: Sunyp01;Achilles_d;fu-hao-huawei;chenweiheng;hulida;tangjie66;zhukkk;SunYaping Merged-by: ascend-robot Description: fix doc bug See merge request: Ascend/torchair!2811	3 个月前

简介

本项目开发了名为TorchAir(Torch Ascend Intermediate Representation)的扩展库，支持用户基于PyTorch框架和TorchNPU插件在昇腾NPU上使用图模式进行推理。 TorchAir继承自PyTorch框架Dynamo模式，将PyTorch的FX图转换为GE计算图，并提供了GE计算图在昇腾NPU的编译与执行的能力。

如果您想了解如何使用TorchAir，可以优先通过访问TorchAir图模式使用指南，了解关于TorchAir的更多信息。

如果您计划为TorchAir贡献代码，请参考README内容。

安装与卸载

1.安装

安装依赖

在安装TorchAir之前，请参考版本配套表和支持型号说明，确保您的硬件能够使用TorchAir，并安装最新昇腾软件栈。

编译准备

克隆TorchAir代码仓

git clone https://gitcode.com/ascend/torchair.git

下载依赖三方库

cd torchair
git submodule update --init --recursive

配置编译环境，执行配置命令

配置只需要进行一次，用于获取pytorch的编译选项（如当前的torch是否开启ABI）及Ascend sdk的安装路径（如果需要在本地CPU上进行调试）。

cd ./torchair
bash ./configure

默认情况下，执行上述命会弹出如下的交互式会话窗口

您的会话可能有所不同。

Please specify the location of python with available torch 2.1.x installed. [Default is /usr/bin/python3]
(You can make this quiet by set env [TARGET_PYTHON_PATH]):

此时，要求您输入安装了 Torch 2.1 版本的python解释器路径，如果默认路径是正确的，直接回车，否则请输入正确的 python 解释器路径。

您可以通过设置 TARGET_PYTHON_PATH 环境变量，来抑制交互式窗口弹出，但是要确保路径是有效的，否则，仍然会要求您输入正确的 python 解释器路径。

键入后，会耗费几秒钟以确保您的输入是有效的，接着，会弹出下面的交互式窗口

Specify the location of ascend sdk for debug on localhost or leave empty.
(You can make this quiet by set env [ASCEND_SDK_PATH]):

如果您不需要在本地CPU模拟执行，可以直接回车跳过。否则，需要输入昇腾处理器开发套件的安装路径（默认安装路径通常为~/Ascend/ascend-toolkit/latest/）。

您可以通过设置 ASCEND_SDK_PATH 环境变量指定 SDK 目录或设置 NO_ASCEND_SDK 环境变量指定不需要 SDK 来抑制交互式窗口弹出。

键入后，等待配置完成。

编译

执行以下命令，编译生成TorchAir安装包：

mkdir build
cd build
cmake ..
make torchair -j8

安装

编译完成后，会在build/dist/dist/目录下生成名为torchair-{version}-py3-none-any.whl的安装包文件。

您可以直接使用pip安装该安装包，或者使用make命令安装至您configure时指定的python环境中。

make install_torchair

2.卸载

torchair的卸载只需要执行命令：

pip3 uninstall torchair

如需要保存卸载日志，可在pip3 uninstall命令后面加上参数--log <PATH>，并对您指定的目录<PATH>做好权限管控。

快速上手

如果您在配置时未指定Ascend sdk的安装路径，则无法执行CPU上的调试，需要在NPU环境上进行测试。

CPU调试时，需要设置LD_LIBRARY_PATH到生成的fake so文件目录以及sdk目录

tools/env.sh会根据配置生成对应的LD_LIBRARY_PATH（如果您在配置时指定了Ascend sdk安装路径）

source tools/env.sh

执行以下python脚本快速验证TorchAir基本功能

import torch
import torchair

config = torchair.CompilerConfig()
npu_backend = torchair.get_npu_backend(compiler_config=config)

class Model(torch.nn.Module):
    def __init__(self):
        super().__init__()

    def forward(self, x, y):
        return torch.add(x, y)

model = Model()
model = torch.compile(model, backend=npu_backend, dynamic=False)
x = torch.randn(2, 2)
y = torch.randn(2, 2)
model(x, y)

特性介绍

TorchAir常用特性如下，全量特性介绍请访问docs获取详细教程。

特性功能	功能介绍
日志功能	日志功能
graph dump功能	图dump功能
dynamo export功能	air格式图导出功能
data dump功能	精度数据dump功能
graph fusion功能	用户自定义关闭/开启部分融合算子功能
experimental性能提升功能	试验性质功能，不同功能适用于特定场景，详见参考资料
converter功能拓展	用户自行扩展模型中缺失的converter功能
支持的aten API清单	支持的aten API清单

版本配套表

本项目与PyTorch、TorchNPU、CANN以及Python的版本匹配详情如下：

TorchAir版本	PyTorch版本	TorchNPU版本	CANN版本	Python版本
master（主线）	2.6.0 +	在研版本	在研版本	Python3.9.x / Python3.10.x / Python3.11.x / Python3.12.x
7.3.0	2.6.0 / 2.7.1 / 2.8.0	7.3.0	8.5.0	Python3.9.x / Python3.10.x / Python3.11.x
7.3.0	2.9.0	7.3.0	8.5.0	Python3.9.x / Python3.10.x / Python3.11.x / Python3.12.x
7.2.0	2.1.0	7.2.0	8.3.RC1	Python3.8.x / Python3.9.x / Python3.10.x / Python3.11.x
7.2.0	2.6.0 / 2.7.1 / 2.8.0	7.2.0	8.3.RC1	Python3.9.x / Python3.10.x / Python3.11.x
7.1.0	2.1.0	7.1.0	8.2.RC1	Python3.8.x / Python3.9.x / Python3.10.x / Python3.11.x
7.1.0	2.5.1 / 2.6.0	7.1.0	8.2.RC1	Python3.9.x / Python3.10.x / Python3.11.x
7.0.0	2.1.0 / 2.3.1 / 2.4.0	7.0.0	8.1.RC1	Python3.8.x / Python3.9.x / Python3.10.x / Python3.11.x
7.0.0	2.5.1	7.0.0	8.1.RC1	Python3.9.x / Python3.10.x / Python3.11.x

支持的型号

Ascend 950PR/Ascend 950DT
Atlas A3 训练系列产品/Atlas A3 推理系列产品
Atlas A2 训练系列产品/Atlas A2 推理系列产品

贡献

如果您计划为TorchAir做出贡献，请参考CONTRIBUTING。

联系我们

安全声明

TorchAir安全声明参考SECURITY_README文件。

许可证

TorchAir插件使用BSD许可证。详见LICENSE文件。

免责声明

致TorchAir使用者

TorchAir提供的模型仅供您用于非商业目的。
对于各模型，TorchAir平台仅提示性地向您建议可用于训练的数据集，华为不提供任何数据集，如您使用这些数据集进行训练，请您特别注意应遵守对应数据集的License，如您因使用数据集而产生侵权纠纷，华为不承担任何责任。
如您在使用TorchAir模型过程中，发现任何问题（包括但不限于功能问题、合规问题），请在GitCode提交issue，我们将及时审视并解决。