1 Issue closed by 1 person

• Question about ncclTopoConnectNodes in topo.cc

  #1498 closed Oct 30, 2024

5 Issues opened by 5 people

• Encounter NCCL error when runing Pytorch example code

  #1504 opened Nov 3, 2024

• Difference between readLL() and readLLFinish() in prims_ll.h

  #1503 opened Oct 31, 2024

• nccl topo about PHB and NODE

  #1502 opened Oct 31, 2024

• Can NCCL and model forward propagation (CUDA matrix operations) be executed simultaneously, and if so, how can this be achieved?

  #1500 opened Oct 30, 2024

• ncclInternalError: Internal check failed

  #1499 opened Oct 30, 2024

11 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• result of sendrecv_perf is wrong

  #1496 commented on Oct 29, 2024 • 0 new comments

• Questions about tuning parameters in graph/tuning.cc

  #1172 commented on Oct 30, 2024 • 0 new comments

• Internal error when submitting a job to a Ray cluster

  #1297 commented on Oct 30, 2024 • 0 new comments

• P2P in non-blocking mode

  #1495 commented on Oct 30, 2024 • 0 new comments

• same data all reduce on H20, but results are different

  #1497 commented on Oct 31, 2024 • 0 new comments

• NCCL_PARAM per ncclComm

  #225 commented on Nov 3, 2024 • 0 new comments

• Is there a design consideration here or just a performance bug?

  #612 commented on Nov 3, 2024 • 0 new comments

• c++/11/bits/std_function.h:435:145: error: parameter packs not expanded with ‘...’: 435 | function(_Functor&& __f)

  #723 commented on Nov 3, 2024 • 0 new comments

• the handle in plugin doesn't work

  #1073 commented on Nov 3, 2024 • 0 new comments

• Add dtype support for int16

  #1078 commented on Nov 3, 2024 • 0 new comments

• The time of alltoall is unstable on two machines

  #1037 commented on Nov 4, 2024 • 0 new comments