Init the main branch for aiter

85762c1a · Xiaowei.zhang · ae0b3521 · 85762c1a · 85762c1a · 85762c1a
Commit 85762c1a authored Apr 15, 2026 by Xiaowei.zhang
20 changed files
--- a/csrc/pybind/rope_general_bwd_pybind.cu
+++ b/csrc/pybind/rope_general_bwd_pybind.cu
--- a/csrc/pybind/rope_general_fwd_pybind.cu
+++ b/csrc/pybind/rope_general_fwd_pybind.cu
--- a/csrc/pybind/rope_pos_fwd_pybind.cu
+++ b/csrc/pybind/rope_pos_fwd_pybind.cu
--- a/csrc/pybind/smoothquant_pybind.cu
+++ b/csrc/pybind/smoothquant_pybind.cu
--- a/csrc/pybind/topk_plain_pybind.cu
+++ b/csrc/pybind/topk_plain_pybind.cu
--- a/csrc/pybind/topk_transform_pybind.cu
+++ b/csrc/pybind/topk_transform_pybind.cu
--- a/csrc/rocm_ops.cpp
+++ b/csrc/rocm_ops.cpp
--- a/das-build.sh
+++ b/das-build.sh
--- a/gen_co.sh
+++ b/gen_co.sh
--- a/gradlib/README.md
+++ b/gradlib/README.md
--- a/gradlib/csrc/grad_funcs.cu
+++ b/gradlib/csrc/grad_funcs.cu
--- a/gradlib/csrc/hipbsolgemm.cu
+++ b/gradlib/csrc/hipbsolgemm.cu
--- a/gradlib/csrc/rocsolgemm.cu
+++ b/gradlib/csrc/rocsolgemm.cu
--- a/gradlib/gradlib/GemmTuner.py
+++ b/gradlib/gradlib/GemmTuner.py
--- a/gradlib/gradlib/gemm_tuner.py
+++ b/gradlib/gradlib/gemm_tuner.py
--- a/gradlib/include/hipbsolgemm.cuh
+++ b/gradlib/include/hipbsolgemm.cuh
--- a/gradlib/include/rocsolgemm.cuh
+++ b/gradlib/include/rocsolgemm.cuh
--- a/gradlib/setup.py
+++ b/gradlib/setup.py
--- a/hsa/gfx936/Cijk_Ailk_Bljk_BBS_BH_UserArgs_MT16x32x32_SN_K1_PGR6_SB1_TT4_2_w4a16.co
+++ b/hsa/gfx936/Cijk_Ailk_Bljk_BBS_BH_UserArgs_MT16x32x32_SN_K1_PGR6_SB1_TT4_2_w4a16.co
--- a/hsa/gfx936/Cijk_Ailk_Bljk_BBS_BH_UserArgs_MT64x128x32_SN_K1_PGR6_SB1_TT4_2_w4a16.co
+++ b/hsa/gfx936/Cijk_Ailk_Bljk_BBS_BH_UserArgs_MT64x128x32_SN_K1_PGR6_SB1_TT4_2_w4a16.co