GitLab

issue/127: Optimize elementwise CUDA code by removing redundancy, change/correct kernel logic when all inputs have the same dtype

issue/127: Refactor ElementwiseInfo, refactor elementwise to use workspace for storing meta, fix misc. issues

issue/127: fix CUDA mix-precision broadcasting input mismatch issue, adjust comment structure and template variable order

enable_if, remove std::move() in elementwise_cpu.h, add <array> inclusion

issue/127: refactor ElementwiseInfo to use utils::Result, change elementwise calcualte and calculateImpl to return infiniStatus_t, add CHECK_CUDA to cuda function calls

issue/127: modify swiglu test to correctly handle broadcast scenarios, add two broadcast testcases, correct elementwise cpu mix-precision implementation

issue/127: refactor elementwise framework, complete CUDA implementation, refactor swiglu using the generic elementwise framework

issue/40: 实现沐曦rms_norm算子

issue/130: GEMM算子CPU平台的omp重构

issue/32: 实现摩尔线程rms_norm算子

issue/111: 添加rmsnorm以及算子编译流程

issue/4 添加cuda causal softmax算子

issue/161 rope和causal softmax支持非原地

issue/111: Merge branch 'main' of github.com:PanZezhong1725/InfiniCore into issue/111-rmsnorm-kunlun

issue/121/feat: 摩尔跟随Result类型修改Gemm构造流程

issue/111: Merge branch 'main' of github.com:PanZezhong1725/InfiniCore into issue/111-rmsnorm-kunlun