Skip to content

Fixing the default value of CCL in infer.py #725

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

Merged
quic-rishinr merged 5 commits into from
Jan 15, 2026
Merged

Fixing the default value of CCL in infer.py #725

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
e9b644d
Evaluating the values of CCL lists for different scenarios
vjanfaza Jan 9, 2026
45c70bd
Evaluating the values of CCL lists for different scenarios
vjanfaza Jan 9, 2026
ec57d5a
Merge branch 'quic:main' into CCL-main
vjanfaza Jan 13, 2026
eab0692
Addressing the default value of CCL lists in infer.py
vjanfaza Jan 13, 2026
29cd6eb
Fixing the default value of CCL in infer.py
vjanfaza Jan 14, 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
14 changes: 12 additions & 2 deletions QEfficient/cloud/infer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -138,6 +138,7 @@ def main(
enable_qnn: Optional[bool] = False,
qnn_config: Optional[str] = None,
trust_remote_code: Optional[bool] = False,
ccl_enabled: Optional[bool] = False,
**kwargs,
) -> None:
"""
Expand Down Expand Up @@ -237,13 +238,16 @@ def main(
if args.mxint8:
logger.warning("mxint8 is going to be deprecated in a future release, use -mxint8_kv_cache instead.")

qaic_config = {"ccl_enabled": True} if ccl_enabled else None

qeff_model = QEFFCommonLoader.from_pretrained(
pretrained_model_name_or_path=model_name,
cache_dir=cache_dir,
hf_token=hf_token,
full_batch_size=full_batch_size,
local_model_dir=local_model_dir,
trust_remote_code=trust_remote_code,
qaic_config=qaic_config,
)

image_path = kwargs.pop("image_path", None)
Expand Down Expand Up @@ -343,15 +347,21 @@ def main(
parser.add_argument(
"--comp-ctx-lengths-prefill",
type=lambda comp_ctx_lengths_prefill: [int(x) for x in comp_ctx_lengths_prefill.split(",")],
default=[512],
default=None,
help="Define ccl list in csv format (e.g.,--comp-ctx-lengths 512,1024,2048).",
)
parser.add_argument(
"--comp-ctx-lengths-decode",
type=lambda comp_ctx_lengths_decode: [int(x) for x in comp_ctx_lengths_decode.split(",")],
default=[2048],
default=None,
help="Define ccl list in csv format (e.g.,--comp-ctx-lengths 512,1024,2048).",
)
parser.add_argument(
"--ccl_enabled",
Copy link
Copy Markdown
Contributor

@quic-rishinr quic-rishinr Jan 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

its better to keep the flag name as enable_ccl

Copy link
Copy Markdown
Contributor Author

@vjanfaza vjanfaza Jan 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We have used this flag name in the ccl implementation. I think it's better to keep it consistent with other scripts.

"--ccl-enabled",
action="store_true",
help="If passed, ccl feature will be activated",
)
parser.add_argument(
"--mxfp6",
"--mxfp6_matmul",
Expand Down
Loading