GitLab

Issue/261: Optimize Torch Implementation of Several Operators

issue/261: optimize the torch implementation of add, causal softmax, gemm, random sample, rearrange, rms norm, rope

issue/36 - Migrate cuda ramdom sample to metax

issue/256 沐曦通信库

issue/238 - Migrate cuda rearrange to metax

issue/39 Migrate cuda causal softmax to metax

issue/37 - Migrate cuda rope to metax

issue/228: swiglu测例0步长添加

Issue/150 添加rope算子测例

issue/142：添加rms_norm算子测例

issue/247 接入昇腾通信库

Issue/141: 沐曦Elementwise以及Swiglu算子实现

Issue/228: infiniop-test框架支持0步长

issue/234 昇腾gemm缓存executor

issue/180：添加clip算子

issue/226: 修改_f32_to_f16()方法内的判断语句条件

feat: 添加昇腾rope算子