DeepSpeed
[CPU] Support SHM based inference_all_reduce in TorchBackend
#5391

Merged

Commits

support shm based allreduce when torchCCL is not installed

delock committed 2 years ago
keep deepspeed.comm.inference_all_reduce interface not changed

delock committed 2 years ago
Merge branch 'master' into gma/gloo_shm_allreduce

loadams committed 2 years ago
fix formatting

delock committed 2 years ago
Merge branch 'master' into gma/gloo_shm_allreduce

tjruwase committed 2 years ago
Change 'SHM' in op builder name into 'ShareMem'

delock committed 2 years ago
add op to inference_all_reduce

delock committed 2 years ago
restore oneccl call parameter

delock committed 2 years ago
Merge branch 'master' into gma/gloo_shm_allreduce

tjruwase committed 2 years ago
Merge branch 'master' into gma/gloo_shm_allreduce

loadams committed 2 years ago