21 Commits (03b32e4eabc5128110ea9d673ba2ffc4bdd9cc2f)

Author SHA1 Message Date
  i-robot 9c6fb2a67e
!31647 clean code 4 years ago
  i-robot 27e417de6a
!31271 Reconstruct the GPU invoking mode 4 years ago
  baihuawei d0f2974316 clean code 4 years ago
  zong-shuai d42daf0dcb reconstruct_gpu_kernel 4 years ago
  hanhuifeng2020 089ce0386a [DynamicShape][Gpu]support dynamic shape for the DCN network 4 years ago
  i-robot c04faad4ad
!30559 modify gpu opt adapter dynamic shape 4 years ago
  changzherui 47ef477fef modify gpu opt adapter dynamic shape 4 years ago
  zhujingxuan 45e7c257af remove duplicate op impl 4 years ago
  ttudu dac6f5425d delete GPU/CPUDynamicKernel 4 years ago
  ckey_Dou e9679ca0bd return default shape when max_shape is empty 4 years ago
  zhoufeng f49b195c39 extract common as an independent shared library 4 years ago
  i-robot ac1463a192
!30298 Package the cuda operators as a dynamic link library 4 years ago
  jinjiali b2d1bad62c package the cuda operations as a dynamic link library 4 years ago
  z00512249 d1e2dfcbd1 add mul dtype register to support bool && uint_8 4 years ago
  zhujingxuan 911e4660ce Rename matrix_triangular_solve to be compatible with tensorflow's MatrixTriangularSolve OP 4 years ago
  yanghaoran 739c4b236a
回退 'Pull Request !30120 : Package the cuda operators as a dynamic link library' 4 years ago
  i-robot 8e21ffae65
!30120 Package the cuda operators as a dynamic link library 4 years ago
  jinjiali cd7db0f57f package the cuda operations as a dynamic link library 4 years ago
  wenbean ba3b65b9af Optimize eigh, backend support 1 or 2 output for GPU/CPU 4 years ago
  wenbean fddedb03c8 Matrix ops move to arrays dir, use scure func 4 years ago
  marui d47e79b04c Refactor ccsrc directories and CMakeLists files 4 years ago