Skip to content

Openvino convertion and quantization #322

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
MattVAD wants to merge 6 commits into
base: master
Choose a base branch
from
Open

Openvino convertion and quantization #322

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
7024d33
add requirements files for deployment
Jul 28, 2025
e78552f
add cv2 in requirements file
Jul 29, 2025
386df9b
openvino.runtime will be removed in version 2026.0
Jul 29, 2025
313cf9a
quantization using nncf
Jul 30, 2025
1c546d6
openvino convertion and quantization
Jul 30, 2025
f5e78bf
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Jul 30, 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
10 changes: 4 additions & 6 deletions .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,4 @@
output/
weight/
*.pyc
wandb/
*.pth
*.onnx
output/
weight/
*.pyc
wandb/
1 change: 1 addition & 0 deletions requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,3 +7,4 @@ scipy
calflops
transformers
loguru
opencv-python
Binary file added single-forever-1024x646.jpg
View file
Open in desktop
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
115 changes: 115 additions & 0 deletions tools/deployment/export_openvino.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,115 @@
"""
D-FINE: Redefine Regression Task of DETRs as Fine-grained Distribution Refinement
Copyright (c) 2024 The D-FINE Authors. All Rights Reserved.
---------------------------------------------------------------------------------
Modified from RT-DETR (https://github.com/lyuwenyu/RT-DETR)
Copyright (c) 2023 lyuwenyu. All Rights Reserved.
"""

import os
import sys

from pathlib import Path

sys.path.insert(0, os.path.join(os.path.dirname(os.path.abspath(__file__)), "../.."))

import openvino as ov
import nncf
import torch
from torchvision import datasets, transforms
import numpy as np


def main(
args,
):
"""main"""
print("Start OpenVINO model convertion...")
if not args.dynamic and args.batch_size == 1:
inp = [1, 3, args.img_size[0], args.img_size[1]]
elif args.batch_size > 1 and args.dynamic:
inp = [-1, 3, -1, -1]
elif args.batch_size > 1:
inp = [-1, 3, args.img_size[0], args.img_size[1]]
elif args.dynamic:
inp = [1, 3, -1, -1]

model = ov.convert_model(
input_model=str(args.model_path),
input=inp,
)
ov.serialize(model, str(Path(args.model_path).with_suffix(".xml")), str(Path(args.model_path).with_suffix(".bin")))
print("OpenVINO model exported")

if args.quantize_nncf :
print("Start NNCF quantization...")

# Instantiate your uncompressed model
model = ov.Core().read_model(str(Path(args.model_path).with_suffix(".xml")))

# # Provide validation part of the dataset to collect statistics needed for the compression algorithm
val_dataset = datasets.ImageFolder(args.calibration_dataset_path, transform=transforms.Compose([transforms.ToTensor()]))
dataset_loader = torch.utils.data.DataLoader(val_dataset, batch_size=1)

# Step 1: Initialize transformation function
def transform_fn(data_item):
images, _ = data_item
orig_size = np.array([args.img_size[0], args.img_size[0]], dtype=np.int64).reshape(
1, 2
)
inputs = {
"images": np.array(images),
"orig_target_sizes": orig_size,
}
return inputs

# Step 2: Initialize NNCF Dataset
calibration_dataset = nncf.Dataset(dataset_loader, transform_fn)
# Step 3: Run the quantization pipeline
quantized_model = nncf.quantize(model, calibration_dataset, preset=nncf.QuantizationPreset.MIXED)
# Step 4: Save model
ov.save_model(quantized_model,str(Path(args.model_path.replace(".onnx","_int8.xml"))))
print("OpenVINO model quantized")


if __name__ == "__main__":
import argparse

parser = argparse.ArgumentParser()
parser.add_argument(
"--model_path",
"-m",
type=str,
required=True,
)
parser.add_argument(
"--dynamic",
"-d",
default=False,
type=bool,
)
parser.add_argument(
"--batch_size",
"-b",
default=1,
type=int,
)
parser.add_argument(
"--img_size",
"-i",
default=(640,640),
type=tuple,
)
parser.add_argument(
"--quantize_nncf",
"-q",
default=False,
type=bool,
)
parser.add_argument(
"--calibration_dataset_path",
"-c",
type=str,
)
args = parser.parse_args()
main(args)
4 changes: 4 additions & 0 deletions tools/deployment/requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,4 @@
onnx
onnxsim
nncf
numpy