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.
 
 
 
 
 
 
mindspore-ci-bot b7b22bb162 !639 unify mutex for model manager 3 years ago
..
analyzer !107 bugfix:fix core question because of DestroyGraphJson lack of return 4 years ago
client remove all msprof, changed to static msprof 3 years ago
common !674 Fix codestyle warnings 3 years ago
engine_manager ge executor support windows 4 years ago
executor remove all msprof, changed to static msprof 3 years ago
ge_local_engine modify log level & rm invalid head files 3 years ago
ge_runtime cpplint magic num & define macro 3 years ago
generator Feature: reset shape of dynamic single op 3 years ago
graph !639 unify mutex for model manager 3 years ago
host_cpu_engine inference supports dynamic shape 3 years ago
host_kernels Fix code check 3 years ago
hybrid change hccl inference 3 years ago
inc update code from yellow zone 20201013 4 years ago
init Remove pass reporter callback to hccl 3 years ago
ir_build !646 Bugfix: format check pref 3 years ago
model test atc dynamic shape 3 years ago
offline Bugfix: single op and ir build question bugfix 3 years ago
omm cpplint cast fix 3 years ago
opskernel_manager Remove pass reporter callback to hccl 3 years ago
plugin/engine remove soft link and support other scene cmake 4 years ago
proto aipp config add related input name and modify proto 4 years ago
session fix cpp lint and errocode 3 years ago
single_op !662 Feature: change profiling open interface 3 years ago
stub profiling AR version 1 3 years ago
CMakeLists.txt Feature: reset shape of dynamic single op 3 years ago
README.md update cmake 4 years ago
ge_inference.mk Feature: reset shape of dynamic single op 3 years ago
ge_runner.mk Feature: reset shape of dynamic single op 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