Skip to content

Add support for compile only backend. #183

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
ktuvw wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Add support for compile only backend. #183

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
7d2813c
Add support for compile only backend. Allows for compilation using AI…
ktuvw Feb 4, 2026
86c8fa5
Fix a line with incorrect naming for "sendnn_compile_only"
ktuvw Feb 4, 2026
81ae14d
Fixed formatting issues
ktuvw Feb 26, 2026
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
15 changes: 13 additions & 2 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 @@ -138,6 +138,13 @@
default="inductor",
choices=["inductor", "eager", "aot_eager"],
)
parser.add_argument(
"--compile_backend_aiu",
type=str,
help="Backend for AIU compilation",
default="sendnn",
choices=["sendnn", "sendnn_compile_only"],
)
parser.add_argument(
"--compile_dynamic",
action="store_true",
Expand Down Expand Up @@ -387,7 +394,10 @@
print("must set AIU_WORLD_RANK_0")
exit()
os.environ.setdefault("FLEX_COMPUTE", "SENTIENT")
os.environ.setdefault("FLEX_DEVICE", "PF")
if args.compile_backend_aiu == "sendnn_compile_only":
os.environ.setdefault("FLEX_DEVICE", "COMPILE")
else:
os.environ.setdefault("FLEX_DEVICE", "PF")

device = torch.device("cpu")
else:
Expand Down Expand Up @@ -597,7 +607,8 @@ def select_int8_module(
fx_config.backed_size_oblivious = "paged" in attn_name
if is_aiu_backend:
model.compile(
backend="sendnn", options={"sendnn.dynamic": args.compile_dynamic_sendnn}
backend=args.compile_backend_aiu,
options={"sendnn.dynamic": args.compile_dynamic_sendnn},
)
else:
# compiling can make first inference pass slow
Expand Down