Skip to content

Introduce Preprocessing for Optimized Quantization in quantize-ort.py #238

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
merged 2 commits into from
Feb 27, 2024
Merged

Introduce Preprocessing for Optimized Quantization in quantize-ort.py #238

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e11a205
Add preprocessing with optimization before quantization
Tim-Siu Feb 26, 2024
212a53d
Clean up unintended changed
Tim-Siu Feb 26, 2024
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
3 changes: 2 additions & 1 deletion tools/quantize/quantize-ort.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@
import onnx
from onnx import version_converter
import onnxruntime
from onnxruntime.quantization import quantize_static, CalibrationDataReader, QuantType, QuantFormat
from onnxruntime.quantization import quantize_static, CalibrationDataReader, QuantType, QuantFormat, quant_pre_process

from transform import Compose, Resize, CenterCrop, Normalize, ColorConvert, HandAlign

Expand Down Expand Up @@ -76,6 +76,7 @@ def check_opset(self):
def run(self):
print('Quantizing {}: act_type {}, wt_type {}'.format(self.model_path, self.act_type, self.wt_type))
new_model_path = self.check_opset()
quant_pre_process(new_model_path, new_model_path)
output_name = '{}_{}.onnx'.format(self.model_path[:-5], self.wt_type)
quantize_static(new_model_path, output_name, self.dr,
quant_format=QuantFormat.QOperator, # start from onnxruntime==1.11.0, quant_format is set to QuantFormat.QDQ by default, which performs fake quantization
Expand Down