Skip to content

Shape inference: ReduceMean dispatcher, quant_pre_process: skip_symbolic_shape bugfix #23558

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
jambayk merged 1 commit into from
Feb 1, 2025
Merged

Shape inference: ReduceMean dispatcher, quant_pre_process: skip_symbolic_shape bugfix #23558

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 6 additions & 0 deletions onnxruntime/python/tools/quantization/shape_inference.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,6 +119,12 @@ def quant_pre_process(
external_names, external_values = extract_raw_data_from_model(input_model)
sess_option.add_external_initializers(list(external_names), list(external_values))
input_model = input_model.SerializeToString()
# the saved optimized model otherwise points to the original external data file name
# which is not available relative to the optimized model file
elif skip_symbolic_shape and save_as_external_data:
sess_option.add_session_config_entry(
"session.optimized_model_external_initializers_file_name", "optimized.onnx.data"
)

sess = onnxruntime.InferenceSession(input_model, sess_option, providers=["CPUExecutionProvider"])
# Close the session to avoid the cleanup error on Windows for temp folders
Expand Down
6 changes: 6 additions & 0 deletions onnxruntime/python/tools/symbolic_shape_infer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -166,6 +166,7 @@ def __init__(self, int_max, auto_merge, guess_output_rank, verbose, prefix=""):
"Range": self._infer_Range,
"Reciprocal": self._pass_on_shape_and_type,
"ReduceSum": self._infer_ReduceSum,
"ReduceMean": self._infer_ReduceMean,
"ReduceProd": self._infer_ReduceProd,
"Reshape": self._infer_Reshape,
"Resize": self._infer_Resize,
Expand Down Expand Up @@ -1603,6 +1604,11 @@ def _infer_ReduceSum(self, node): # noqa: N802
)
)

def _infer_ReduceMean(self, node): # noqa: N802
if get_opset(self.out_mp_) >= 18:
# reduce mean spec 18+ is same as reduce sum spec 13+
self._infer_ReduceSum(node)

def _infer_ReduceProd(self, node): # noqa: N802
axes = get_attribute(node, "axes")
keep_dims = get_attribute(node, "keepdims", 1)
Expand Down