Skip to content
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

add op_types_to_quantize to get_qnn_qdq_config #23458

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
Open

add op_types_to_quantize to get_qnn_qdq_config #23458

Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
49f1f7e
add op_types_to_quantize to get_qnn_qdq_config
Jan 22, 2025
528aefe
use set
Jan 24, 2025
f79f00a
nit
Jan 24, 2025
417dab3
lintrunner -a
Jan 26, 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
8 changes: 7 additions & 1 deletion onnxruntime/python/tools/quantization/execution_providers/qnn/quant_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,7 @@ def get_qnn_qdq_config(
weight_symmetric: bool | None = None,
keep_removable_activations: bool = False,
stride: int | None = None,
op_types_to_quantize: list[str] | None = None,
) -> StaticQuantConfig:
"""
Returns a static quantization configuration suitable for running QDQ models on QNN EP.
Expand Down Expand Up @@ -117,6 +118,7 @@ def get_qnn_qdq_config(
are automatically removed if activations are asymmetrically quantized. Keeping these activations
is necessary if optimizations or EP transformations will later remove
QuantizeLinear/DequantizeLinear operators from the model.
op_types_to_quantize: If set to None, all operator types will be quantized except for OP_TYPES_TO_EXCLUDE

Returns:
A StaticQuantConfig object
Expand Down Expand Up @@ -161,7 +163,11 @@ def get_qnn_qdq_config(
name_to_initializer,
)

op_types_to_quantize_set = set(op_types_to_quantize) if op_types_to_quantize else None

for node in model.graph.node:
if op_types_to_quantize_set and node.op_type not in op_types_to_quantize_set:
continue
op_types.add(node.op_type)
qnn_compat.process_node(node)

Expand Down Expand Up @@ -189,7 +195,7 @@ def get_qnn_qdq_config(
calibrate_method=calibrate_method,
activation_type=activation_type,
weight_type=weight_type,
op_types_to_quantize=list(op_types.difference(OP_TYPES_TO_EXCLUDE)),
op_types_to_quantize=op_types_to_quantize if op_types_to_quantize else list(op_types.difference(OP_TYPES_TO_EXCLUDE)),
per_channel=per_channel,
use_external_data_format=(model_has_external_data or model.ByteSize() >= MODEL_SIZE_THRESHOLD),
extra_options=extra_options,
Expand Down
Loading