singa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <>
Subject [GitHub] [singa-doc] chrishkchris commented on issue #14: rearrange contents in
Date Sat, 04 Apr 2020 08:31:10 GMT
chrishkchris commented on issue #14: rearrange contents in
   > Is nccl_id passed to `train_mnist_cnn` for training with MPI?
   No, in the case of MPI, nccl_id is generated and immediately broadcasted by MPI from rank0
to every rank. 
     if (MPIRankInGlobal == 0) ncclGetUniqueId(&id);
     MPICHECK(MPI_Bcast((void *)&id, sizeof(id), MPI_BYTE, 0, MPI_COMM_WORLD));
   > For multiprocessing, num_gpu = gpu_per_node, hence we only need num_gpu?
    This is reserve for future use, because previously I planned using socket to broadcast
ncclid instead of MPI

This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:

With regards,
Apache Git Services

View raw message