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.
 
 
 
 
 
 
lianghuikang f595a577dd add op_precision_mode option and support op_debug_level = 4 3 years ago
..
analyzer Normalize include files... 3 years ago
client Normalize include files... 3 years ago
common !1781 fix cmetric 3 years ago
engine_manager Normalize include files... 3 years ago
executor Normalize include files... 3 years ago
ge_local_engine Normalize include files... 3 years ago
ge_opt_info opt_info 3 years ago
ge_runtime Normalize include files... 3 years ago
generator Normalize include files... 3 years ago
graph !1840 opt info 3 years ago
host_kernels Normalize include files... 3 years ago
hybrid !1821 Create NodeExecute on-demand 3 years ago
inc Normalize include files... 3 years ago
init Normalize include files... 3 years ago
ir_build add op_precision_mode option and support op_debug_level = 4 3 years ago
model 回退 'Pull Request !1704 : remove updation of session_id' 3 years ago
offline add op_precision_mode option and support op_debug_level = 4 3 years ago
opskernel_manager Normalize include files... 3 years ago
plugin/engine Normalize include files... 3 years ago
session add op_precision_mode option and support op_debug_level = 4 3 years ago
single_op !1804 Aicpu single_op always dynamic execute. 3 years ago
stub python3,7 modify python3 3 years ago
CMakeLists.txt opt_info 3 years ago
README.md update cmake 4 years ago
ge_inference.mk !1564 Bugfix: Delete abandoned optimize pass 3 years ago
ge_runner.mk !1564 Bugfix: Delete abandoned optimize pass 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