BladeDISC icon indicating copy to clipboard operation
BladeDISC copied to clipboard

Published 20 hours ago • verified publisher icon alibaba

support collective operators

Open Yancey1989 opened this issue 11 months ago • 0 comments

To optimize distributed training graph (DP, FSDP), DISC needs to support collective ops as a preliminary preparation

  • [ ] support collective ops compilation and execution (all_reduce, all_gather, reduce_scatter) @Yancey1989
    • #1275
    • #1288
  • [ ] support async collective ops execution @eedalong
    • #1287

Yancey1989 avatar Mar 15 '24 07:03 Yancey1989