Skip to content

numa support for inference.py #158

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

numa support for inference.py #158

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
75b5096
numa support
mserranos Oct 24, 2025
25cb3d3
numa set local alloc
mserranos Oct 25, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 21 additions & 0 deletions aiu_fms_testing_utils/scripts/inference.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -257,6 +257,12 @@
default=0,
help="Timeout to use for messaging in minutes. Default set by PyTorch dist.init_process_group",
)
parser.add_argument(
"--numa",
action="store_true",
help="NUMA aware task distribution (requires distributed option)",
)

args = parser.parse_args()

attention_map = {
Expand Down Expand Up @@ -327,6 +333,21 @@
dist.init_process_group()
# Fix until PT 2.3
torch._C._distributed_c10d._register_process_group("default", dist.group.WORLD)
if args.numa:
try:
from numa import info
numa_num_nodes = info.get_num_configured_nodes()
numa_world_size = dist.get_world_size()
numa_size_per_node = numa_world_size // numa_num_nodes
from numa import schedule
numa_rank = dist.get_rank()
numa_node = dist.get_rank() // numa_size_per_node
schedule.run_on_nodes(numa_node)
from numa import memory
memory.set_local_alloc()
dprint(f"NUMA: process {numa_rank} set to node {numa_node}")
except:
dprint(f"NUMA not available in this machine, please install libnuma libraries")
aiu_setup.aiu_dist_setup(dist.get_rank(), dist.get_world_size())

if args.device_type == "cuda":
Expand Down