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.
3165 Commits
28 Branches
21 MB
2351 Download
Tree: 8eb49d2da3
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '8eb49d2da3'
${ noResults }
graphengine/third_party/fwkacllib/inc/ops/OWNERS

OWNERS 755 B
Raw Normal View History

upgrade Ascend package 11 Feb 22
3 years ago
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
upgarde Ascend software package 1 Sep 22
2 years ago
 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960 
  1. approvers:

  2. - gegenhua

  3. - qiaohairong

  4. reviewers:

  5. - chuqingxi

  6. - wang-jintang

  7. - luanma_bl

  8. - chen-kang30

  9. - li-xulong

  10. - Allan_Yu

  11. - minshen

  12. - pan-jixing

  13. - yl_wang

  14. - lijie176

  15. - mabing726

  16. - miao-fangzheng

  17. - huang-qiang002

  18. - su-yueming

  19. - chenpeng-hw

  20. - wang_jianle

  21. - luanma_bl

  22. - LDLD0524

  23. - wywismygod2020

  24. - koala-zhang

  25. - zhu-jingjing

  26. - zhaozhihui5

  27. - simbaliuxx

  28. - lyxyz

  29. - zhou-qilong

  30. - block0219

  31. - hanfuwei

  32. - xchu42

  33. - sheng-nan

  34. - yangjing88

  35. - alexlak

  36. - xig514

  37. - jellylj

  38. - brightlyking

  39. - liuzhenyuhw

  40. - djh602

  41. - wangjiangben_hw

  42. - li1jie

  43. - liujun2014

  44. - soupkey

  45. - wu-shengji

  46. - cimeng

  47. - granpad7

  48. - tc1qaz

  49. - Ronnie_zheng

  50. - xiexianhu

  51. - zhaoping12

  52. - tanshengshun

  53. - fanqirui

  54. - xu-binglin

  55. - yangyang016

  56. - zhangzhongzt

  57. - gegenhua

  58. - qiaohairong

  59. options:

  60.   no_parent_owners: true

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