| Files | Last commit | Last update |
|---|---|---|
docs: add English translations for operator READMEs and repo docs Co-authored-by: jenny_xiang<jenny.xiang@huawei.com> # message auto-generated for no-merge-commit merge: !13351 merge docs/add-english-docs into master docs: add English translations for operator READMEs and repo docs Created-by: jenny_xiang Commit-by: jenny_xiang Merged-by: cann-robot Description: ## Description Add English translations for 652 markdown documents: - Repo-level guides: README_en.md, CONTRIBUTING_en.md, SECURITY_en.md, CHANGELOG_en.md, codearts_check_rule_config_en.md - Operator READMEs under activation/, conv/, control/, foreach/, hash/, index/, loss/, matmul/, norm/, optim/, pooling/, quant/, rnn/, vfusion/, etc. - API and design docs under docs/ (aclnn_*, ascendir_*, fusion pass docs) - examples/ and torch_extension/ guides Note: a few cross-reference links whose English targets are not published yet temporarily point to their Chinese counterparts to keep links valid; they will be switched back to the English docs in a follow-up PR after this one is merged. ## Related Issue N/A ## Test - Verified all 1,344 intra-repo markdown links in the added files resolve against master: 0 broken links, 0 missing targets. ## Documentation Update - This PR only adds English documentation files (*_en.md); no existing Chinese documents are modified. ## Type Labels - [ ] Bug fix - [ ] New feature - [ ] Performance optimization - [x] Documentation update - [ ] Other, please describe: ## AI/Agent-Generated Declaration - [x] AI-assisted authoring See merge request: cann/ops-nn!13351 | 18 days ago | |
新增 assert_async 算子 Co-authored-by: ASCEND222<dongfei16@h-partners.com> # message auto-generated for no-merge-commit merge: !12521 merge master into master 新增 assert_async 算子 Created-by: ASCEND222 Commit-by: ASCEND222 Merged-by: cann-robot Description: ## 描述 <!--在这里详细描述你的改动,包括改动的原因和所采取的方法。--> 新增 assert_async 算子 - 算子功能:断言输入张量 input_condition(要求 numel == 1)的数值非 0。若其数值为 0(含 bool 的 False), 设备侧打印属性 assert_msg 指定的消息并触发算子失败;数值非 0 直接通过。本算子**无输出张量**, 语义对齐 PyTorch 的 torch._assert_async(含 .msg 重载语义)。 - 判定口径(bit-exact,无数值计算): - 浮点类型(FLOAT16、FLOAT、DOUBLE、BFLOAT16)按符号位掩码判定:FLOAT 取 & 0x7FFFFFFF,FLOAT16/BFLOAT16 取 & 0x7FFF, DOUBLE 取 & 0x7FFFFFFFFFFFFFFF。+0.0 与 -0.0 均判失败,NaN、+inf、-inf 判通过。 - 复数类型(COMPLEX64、COMPLEX128)实部与虚部精确同时为 0(含 -0.0 分量)判失败;NaN/±inf 分量通过。 - 整型与布尔类型判全字节零:0(含 bool 的 False)判失败,非 0(含 True)判通过。 - 失败通路:kernel 内以 AscendC::printf 打印消息后调用 AscendC::Trap() 中断 AI Core,使算子执行失败; 失败状态经 stream 同步(如 aclrtSynchronizeStream)回传给上层。判定通过时不产生任何打印或输出。 ## 参数说明 | 参数名 | 输入/输出/属性 | 描述 | 数据类型 | 数据格式 | |--------|--------------|------|---------|---------| | input_condition | 输入 | 待断言的条件张量,要求 numel == 1(rank-0 或任意收缩为单元素的 shape)。数值为 0(含 bool 的 False)判失败。 | BOOL、INT8、INT16、INT32、INT64、UINT8、FLOAT16、FLOAT、DOUBLE、BFLOAT16、COMPLEX64、COMPLEX128 | ND | | assert_msg | 可选属性 | 断言失败时打印的消息字符串,默认 ""。可用长度 254 字节(长度须小于 255;缓冲 256 字节,含结尾 '\0'),超长时返回 ACLNN_ERR_PARAM_INVALID(161002)。 | String | - | ## 关联的Issue <!-- 如果这个PR是为了解决特定的Issue,请在这里提供Issue链接。例如:关联Issue #000--> https://gitcode.com/cann/ops-nn/issues/7330 ## 测试 <!--描述进行了哪些测试来验证你的改动。--> ## 文档更新 <!--如果这个PR包含文档的更新,请在这里指出。例如:更新了README.md文件。--> ## 类型标签 <!-- [x] 表示选中 --> - [ ] Bug修复 - [x] 新特性 - [ ] 性能优化 - [ ] 文档更新 - [ ] 其他,请描述: ## AI/Agent生成声明 <!-- [x] 表示选中 --> - [x] AI辅助编写 See merge request: cann/ops-nn!12521 | 10 days ago | |
fix(npu_alloc_float_status): 算子质量加固——kernel块写改造/tiling负向UT补齐/文档约束修正 Co-authored-by: dengguojie<dengguojie@huawei.com> # message auto-generated for no-merge-commit merge: !13891 merge hardening/NPUAllocFloatStatus into master fix(npu_alloc_float_status): 算子质量加固——kernel块写改造/tiling负向UT补齐/文档约束修正 Created-by: dengguojie Commit-by: dengguojie Merged-by: cann-robot Description: ## 描述 NPUAllocFloatStatus(control/npu_alloc_float_status,Ascend 950/Arch35)质量加固修复,共 21 项修复明细见汇总 issue。 **修复要点**: - **Kernel 输出路径块写改造(核心)**: 循环 SetValue 逐元素写 GM 改为 TPipe+Duplicate+DataCopy 32B 块写,补 V→MTE3 事件屏障(修复 MTE3 先于向量单元读 UB 的搬出残留)。 - **Tiling UT 修复与负向补齐**: UT 文件名对齐收集规则;哑输入声明适配 CANN 9.3.0(拒绝 input_num=0);新增 5 个 tiling 校验负向用例。 - **文档**: README 中英补域外输入拒绝条件与算子 IR 链接,删零信息量伪约束;op_list.md aclnn 列 ✓→✗(无 aclnn 交付)。 - **样例**: 宿主缓冲泄漏补 delete[];废弃头迁移(array_ops.h/tensor.h→ops_proto_math.h,ge_error_codes.h→error_codes.h)。 - **规范化**: kernel 入口去 _apt 后缀,def.cpp opFile 配置行删除(对齐 kernel 配置项指南);CMakeLists 许可头对齐;proto.h 注释对齐。 - **测试资产**: golden __spec__ 补 GE/TF 键并 lazy import;新增 kernel/geir/tf 三通路看护集 CSV。 ## 关联的Issue Closes #8377 ## 测试 - kernel 通路 binary_equal 全绿(块写改造后逐位一致);UT 7/7 PASS(含 5 负向),tiling 覆盖率 94.2%→96.2%。 - 看护集 CSV 经 TTK validate 与 header 校验全绿;pre-commit(含 OAT)对变更文件定向全绿。 ## 文档更新 - README.md / README_en.md(中英同步)与 docs/zh/op_list.md,详见修复要点。 ## 类型标签 <!-- [x] 表示选中 --> - [x] Bug修复 - [ ] 新特性 - [ ] 性能优化 - [x] 文档更新 - [ ] 其他,请描述: ## AI/Agent生成声明 <!-- [x] 表示选中 --> - [x] AI辅助编写 See merge request: cann/ops-nn!13891 | 12 days ago | |
fix: 加固npu_clear_float_status与sigmoid_cross_entropy_with_logits_grad两个长尾算子 Co-authored-by: klein8793<zhangli24@huawei.com> # message auto-generated for no-merge-commit merge: !14069 merge dev_zl_2026_1106 into master fix: 加固npu_clear_float_status与sigmoid_cross_entropy_with_logits_grad两个长尾算子 Created-by: klein8793 Commit-by: klein8793 Merged-by: cann-robot Description: ## 描述 对两个长尾算子进行加固,修复潜在缺陷并提升健壮性: **npu_clear_float_status** - 将 VEC_DUP_SIZE/VEC_DUP_SIZE_BYTES 常量从 op_kernel/arch35/npu_clear_float_status_simt.h 迁移到 op_kernel/arch35/npu_clear_float_status_tiling_data.h,Host Tiling 与 Kernel 引用同一常量(单一来源),避免两侧各自维护独立常量导致不一致;tiling_arch35.cpp 中的 TBUF_REQUIRED_SIZE 改为引用 VEC_DUP_SIZE_BYTES,并同步更新 static_assert。 - 测试 golden.py 由 TensorFlow 实现切换为 Torch 实现。 **sigmoid_cross_entropy_with_logits_grad** - 样例 test_geir_sigmoid_cross_entropy_with_logits_grad.cpp:GenData 改用 std::vector<std::vector<float>> 管理输入缓冲,消除 new[] 未释放的内存泄漏;AddGraph 增加失败检查;补齐 GEFinalize 成功路径与失败路径的 delete session。 - sigmoid_cross_entropy_with_logits_grad_arch35_tiling.cpp:totalElements 累乘增加负维度检查与 int64 溢出检查。 - sigmoid_cross_entropy_with_logits_grad_proto.h:dout 的 shape 约束由"满足广播关系"修正为"元素总数与 predict 相同,不支持广播"。 - README:参数名 d_out 统一为 dout,与 proto 及代码保持一致。 - 新增 tests/assets/golden.py 精度标杆。 ## 关联的Issue - #8855 ## 测试 冒烟测试已通过,算子L0级测试已通过 ## 文档更新 - 更新 loss/sigmoid_cross_entropy_with_logits_grad/README.md(参数名与 shape 约束说明)。 ## 类型标签 - [x] Bug修复 - [ ] 新特性 - [ ] 性能优化 - [ ] 文档更新 - [ ] 其他,请描述: ## AI/Agent生成声明 - [ ] AI辅助编写 See merge request: cann/ops-nn!14069 | 6 days ago | |
docs: add English translations for operator READMEs and repo docs Co-authored-by: jenny_xiang<jenny.xiang@huawei.com> # message auto-generated for no-merge-commit merge: !13351 merge docs/add-english-docs into master docs: add English translations for operator READMEs and repo docs Created-by: jenny_xiang Commit-by: jenny_xiang Merged-by: cann-robot Description: ## Description Add English translations for 652 markdown documents: - Repo-level guides: README_en.md, CONTRIBUTING_en.md, SECURITY_en.md, CHANGELOG_en.md, codearts_check_rule_config_en.md - Operator READMEs under activation/, conv/, control/, foreach/, hash/, index/, loss/, matmul/, norm/, optim/, pooling/, quant/, rnn/, vfusion/, etc. - API and design docs under docs/ (aclnn_*, ascendir_*, fusion pass docs) - examples/ and torch_extension/ guides Note: a few cross-reference links whose English targets are not published yet temporarily point to their Chinese counterparts to keep links valid; they will be switched back to the English docs in a follow-up PR after this one is merged. ## Related Issue N/A ## Test - Verified all 1,344 intra-repo markdown links in the added files resolve against master: 0 broken links, 0 missing targets. ## Documentation Update - This PR only adds English documentation files (*_en.md); no existing Chinese documents are modified. ## Type Labels - [ ] Bug fix - [ ] New feature - [ ] Performance optimization - [x] Documentation update - [ ] Other, please describe: ## AI/Agent-Generated Declaration - [x] AI-assisted authoring See merge request: cann/ops-nn!13351 | 18 days ago | |
docs: add English translations for operator READMEs and repo docs Co-authored-by: jenny_xiang<jenny.xiang@huawei.com> # message auto-generated for no-merge-commit merge: !13351 merge docs/add-english-docs into master docs: add English translations for operator READMEs and repo docs Created-by: jenny_xiang Commit-by: jenny_xiang Merged-by: cann-robot Description: ## Description Add English translations for 652 markdown documents: - Repo-level guides: README_en.md, CONTRIBUTING_en.md, SECURITY_en.md, CHANGELOG_en.md, codearts_check_rule_config_en.md - Operator READMEs under activation/, conv/, control/, foreach/, hash/, index/, loss/, matmul/, norm/, optim/, pooling/, quant/, rnn/, vfusion/, etc. - API and design docs under docs/ (aclnn_*, ascendir_*, fusion pass docs) - examples/ and torch_extension/ guides Note: a few cross-reference links whose English targets are not published yet temporarily point to their Chinese counterparts to keep links valid; they will be switched back to the English docs in a follow-up PR after this one is merged. ## Related Issue N/A ## Test - Verified all 1,344 intra-repo markdown links in the added files resolve against master: 0 broken links, 0 missing targets. ## Documentation Update - This PR only adds English documentation files (*_en.md); no existing Chinese documents are modified. ## Type Labels - [ ] Bug fix - [ ] New feature - [ ] Performance optimization - [x] Documentation update - [ ] Other, please describe: ## AI/Agent-Generated Declaration - [x] AI-assisted authoring See merge request: cann/ops-nn!13351 | 18 days ago | |
docs: add English translations for operator READMEs and repo docs Co-authored-by: jenny_xiang<jenny.xiang@huawei.com> # message auto-generated for no-merge-commit merge: !13351 merge docs/add-english-docs into master docs: add English translations for operator READMEs and repo docs Created-by: jenny_xiang Commit-by: jenny_xiang Merged-by: cann-robot Description: ## Description Add English translations for 652 markdown documents: - Repo-level guides: README_en.md, CONTRIBUTING_en.md, SECURITY_en.md, CHANGELOG_en.md, codearts_check_rule_config_en.md - Operator READMEs under activation/, conv/, control/, foreach/, hash/, index/, loss/, matmul/, norm/, optim/, pooling/, quant/, rnn/, vfusion/, etc. - API and design docs under docs/ (aclnn_*, ascendir_*, fusion pass docs) - examples/ and torch_extension/ guides Note: a few cross-reference links whose English targets are not published yet temporarily point to their Chinese counterparts to keep links valid; they will be switched back to the English docs in a follow-up PR after this one is merged. ## Related Issue N/A ## Test - Verified all 1,344 intra-repo markdown links in the added files resolve against master: 0 broken links, 0 missing targets. ## Documentation Update - This PR only adds English documentation files (*_en.md); no existing Chinese documents are modified. ## Type Labels - [ ] Bug fix - [ ] New feature - [ ] Performance optimization - [x] Documentation update - [ ] Other, please describe: ## AI/Agent-Generated Declaration - [x] AI-assisted authoring See merge request: cann/ops-nn!13351 | 18 days ago | |
init | 1 year ago |
| Files | Last commit | Last update |
|---|---|---|
| 18 days ago | ||
| 10 days ago | ||
| 12 days ago | ||
| 6 days ago | ||
| 18 days ago | ||
| 18 days ago | ||
| 18 days ago | ||
| 1 year ago |