Commits · ea66e5e5a2a9a7c005e09aaa6c0a97a352db364a · OpenDAS / FastMoE

03 Feb, 2021 1 commit
- fix ensure device index bug · ea66e5e5
  Rick Ho authored Feb 03, 2021
  
  ea66e5e5
01 Feb, 2021 1 commit
- swap local scatter and gather kernel functions · d2039fc7
  Rick Ho authored Feb 01, 2021
  
  d2039fc7
29 Jan, 2021 1 commit
- scatter gather kernel support for non-equal shapes and fix tests · 72e9bc9e
  Rick Ho authored Jan 29, 2021
  
  72e9bc9e
28 Jan, 2021 4 commits
- support fp16 · 49c97411
  Rick Ho authored Jan 28, 2021
  
  49c97411
- make test run on nccl version, but fails in correctness · 2d250fbf
  Rick Ho authored Jan 28, 2021
  
  2d250fbf
- hack nccl of pytorch · 293eef6d
  Rick Ho authored Jan 28, 2021
  
  293eef6d
- single node use torch cuda expert count · a526f438
  Rick Ho authored Jan 28, 2021
  
  a526f438
25 Jan, 2021 2 commits
- remove redundant files · 241cdb78
  Rick Ho authored Jan 25, 2021
  
  241cdb78
- move original files to examples · 0c47afbb
  Rick Ho authored Jan 25, 2021
  
  0c47afbb