You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
l00444296 19b031f2a3 Feature:Support open ir interface in fwkacllib package 3 years ago
..
analyzer !107 bugfix:fix core question because of DestroyGraphJson lack of return 3 years ago
client modified: cmake/external_libs/gflags.cmake 3 years ago
common The model subscription function of the profile module is added to ge. 3 years ago
engine_manager update cmake 4 years ago
executor The model subscription function of the profile module is added to ge. 3 years ago
ge_local_engine modified: cmake/external_libs/gflags.cmake 3 years ago
ge_runtime udapte tash proto 3 years ago
generator acl unregistered op error 3 years ago
graph !295 dts:error message add 3 years ago
host_cpu_engine modified: cmake/external_libs/gflags.cmake 3 years ago
host_kernels slice tensor no originshape 3 years ago
hybrid !261 profiling subscribe 3 years ago
inc update code from yellow zone 20201013 3 years ago
init V200 support NAN INF 3 years ago
ir_build fix aclInferShapeAndType 3 years ago
model update newest yellow zone code 20201020 3 years ago
offline delete atc memory limit 3 years ago
omm update ge folder 4 years ago
opskernel_manager modified: ge/opskernel_manager/ops_kernel_builder_manager.cc 3 years ago
plugin/engine update blue include behind the yellow zone 3 years ago
proto aipp config add related input name and modify proto 3 years ago
session Feature:Support user options of aclgrphParse interface 3 years ago
single_op !196 fix dynamic shape with while 3 years ago
stub update code from yellow zone 20201013 3 years ago
CMakeLists.txt Feature:Support open ir interface in fwkacllib package 3 years ago
README.md update cmake 4 years ago
ge_inference.mk modified: cmake/external_libs/gflags.cmake 3 years ago
ge_runner.mk Feature:Support open ir interface in fwkacllib package 3 years ago
module.mk update ge folder 4 years ago

图引擎模块(GE)是MindSpore的一个子模块,其代码由C++实现,位于前端模块ME和底层硬件之间,起到承接作用。图引擎模块以ME下发的图作为输入,然后进行一系列的深度图优化操作,最后输出一张可以在底层硬件上高效运行的图。GE针对昇腾AI处理器的硬件结构特点,做了特定的优化工作,以此来充分发挥出昇腾AI处理器的强大算力。在进行模型训练/推理时,GE会被自动调用而用户并不感知。GE主要由GE API和GE Core两部分组成,详细的架构图如下所示

C++ Text Makefile Shell CMake other