| feat: 批量增加 34 个 arch35 算子 Co-authored-by: ChenCheng<chencheng324@huawei.com> # message auto-generated for no-merge-commit merge: !7 merge feat/ops-batch-arch35 into master feat: 批量增加 34 个 arch35 算子 Created-by: cc2138 Commit-by: ChenCheng Merged-by: cann-robot Description: ## 变更描述 / Description 本 PR 批量新增 34 个目标架构 arch35(Ascend 950 系列)的算子实现与测试,并更新接口支持清单与接口文档,主要改动如下: ### 1. 新增 34 个算子实现(applications/llm/<op>/arch35/) - 特殊数学函数:bessel_j0 / bessel_j1 / bessel_y0 / bessel_y1 / shifted_chebyshev_polynomial_t / shifted_chebyshev_polynomial_u / shifted_chebyshev_polynomial_v / shifted_chebyshev_polynomial_w / digamma / polygamma / erfcx / entr / logndtr - 归约与排序:cummax / cumprod / kthvalue / mode - 随机采样:cauchy / log_normal - 量化:int_repr / make_per_tensor_quantized - 池化 / 卷积 / 插值:adaptive_avg_pool2d / depthwise_conv3d / upsample_trilinear3d - 损失:multi_margin_loss / multilabel_margin_loss - 索引 / 查找:index_reduce / searchsorted / tril_indices / triu_indices - 其它:angle / complex / conjphysical / foreach_ceil - 每个算子含 <op>.asc(Schema 注册 + Meta 函数 + Kernel 实现 + NPU Dispatch 四段式)与 CMakeLists.txt(add_sources()) ### 2. 新增算子测试(tests/<op>/) - 每个算子含 test_<op>.py:与 PyTorch CPU 参考实现对比的精度测试,覆盖各算子支持的数据类型、shape/dim 组合、边界与非法输入拒绝 ### 3. 更新接口文档 - docs/zh/op_support_list.md:接口支持清单新增上述 34 个算子(ascend950) - docs/zh/api_list.md:新增上述 34 个算子的接口说明 ## 改动类型 / Change Type - [x] Bug 修复 / Bug Fix - [x] 新功能 / New Feature - [ ] 性能优化 / Performance - [x] 代码重构 / Refactoring - [x] 文档更新 / Documentation - [x] 测试相关 / Test - [ ] 其它 / Other ## 关联 Issue / Related Issues - Closes #5 - References #5 ## 测试信息 / Testing - [ ] 单元测试通过 / UT passed - [x] 集成测试通过 / ST passed - [x] 人工验证通过 / Manual verified ## 检查清单 / Checklist - [x] 代码符合规范 / Code follows style guide - [x] 测试添加并通过 / Tests added and passed - [x] 文档已更新 / Docs updated if needed - [x] 无硬编码敏感信息 / No secrets hardcoded - [x] 提交信息符合规范 / Commit message follows convention See merge request: cann/ops-multimodal-fusion!7 | 4 天前 |