hummingbird
/
graphengine
Not watched
2
0
Fork 0
Code
Releases 13 Wiki Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain
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.
1997 Commits
28 Branches
21 MB
0 Download
 
 
 
 
 
 
Tree: 9cda3adfb4
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '9cda3adfb4'
${ noResults }
graphengine/ge/graph/load/model_manager
mindspore-ci-bot 649f555801 !1347 Performance optimization 3 years ago
..
task_info rm op_desc 3 years ago
aipp_utils.cc Rename new_model_manager to model_manager. 3 years ago
aipp_utils.h Rename new_model_manager to model_manager. 3 years ago
cpu_queue_schedule.cc fix mdc mbuf_list order error 3 years ago
cpu_queue_schedule.h fix mdc mbuf_list order error 3 years ago
data_dumper.cc modify data dumper 3 years ago
data_dumper.h modify data dumper 3 years ago
data_inputer.cc Rename new_model_manager to model_manager. 3 years ago
data_inputer.h Rename new_model_manager to model_manager. 3 years ago
davinci_model.cc Performance optimization 3 years ago
davinci_model.h Unique label goto addr 3 years ago
model_manager.cc fix bug of dynamic shape load error 3 years ago
model_manager.h fix bug of dynamic shape load error 3 years ago
model_utils.cc static check 3 years ago
model_utils.h Rename new_model_manager to model_manager. 3 years ago
tbe_handle_store.cc Rename new_model_manager to model_manager. 3 years ago
tbe_handle_store.h Rename new_model_manager to model_manager. 3 years ago
ts_mem_mall.h hide character 3 years ago
zero_copy_offset.cc fix mdc mbuf_list order error 3 years ago
zero_copy_offset.h Performance optimization 3 years ago
zero_copy_task.cc Rename new_model_manager to model_manager. 3 years ago
zero_copy_task.h Rename new_model_manager to model_manager. 3 years ago
Delete task

Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.

cancel
confirm

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