OnnxSlim support

[inisis]

Search before asking

• I have searched the YOLOv8 issues and found no similar feature requests.

Description

Hi, we have developed a tool called onnxslim, which can help slim exported onnx model, and it's pure python, and it works well on yolo, should we intergrate it into awesome ultralytics.

Use case

when export to onnx, we can use onnxslim to slim our exported model

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[github-actions]

👋 Hello @inisis, thank you for your interest in Ultralytics YOLOv8 🚀! We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Join the vibrant Ultralytics Discord 🎧 community for real-time conversations and collaborations. This platform offers a perfect space to inquire, showcase your work, and connect with fellow Ultralytics users.

Install

Pip install the ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8.

pip install ultralytics

Environments

YOLOv8 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLOv8 Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[inisis]

@glenn-jocher Can you take a look, I have tested on yolov8n model, and it works really good for reducing number of operators.

[Kayzwer]

@inisis does OnnxSlim also speed up for TensorRT (since model have to be convert to onnx first) ?

[inisis]

@inisis does OnnxSlim also speed up for TensorRT (since model have to be convert to onnx first) ?

I do believe so, onnxslim depends on nv onnx-graphsurgeon, but have better perfromance.

[Kayzwer]

@inisis does OnnxSlim also speed up for TensorRT (since model have to be convert to onnx first) ?

I do believe so, onnxslim depends on nv onnx-graphsurgeon, but have better perfromance.

does it works with onnx-simplify?

[Kayzwer]

@inisis I will make a pull request to apply this in the export function

[Kayzwer]

@inisis pull request for the integration, thanks for the tool :)

[inisis]

@Kayzwer Hi, CI failed, but I think it's not a problem with onnxslim, and how can I compare the performance with/without onnxslim, and one thing to mention, onnxslim can be used in arm devices.

[Kayzwer]

@Kayzwer Hi, CI failed, but I think it's not a problem with onnxslim, and how can I compare the performance with/without onnxslim, and one thing to mention, onnxslim can be used in arm devices.

from ultralytics import YOLO

model = YOLO("yolov8n.pt")
model.export(format="onnx", slim=True) # slimmed

model.export(format="onnx")

[Kayzwer]

@Kayzwer Hi, CI failed, but I think it's not a problem with onnxslim, and how can I compare the performance with/without onnxslim, and one thing to mention, onnxslim can be used in arm devices.

from ultralytics import YOLO

model = YOLO("yolov8n.pt")
model.export(format="onnx", slim=True) # slimmed

model.export(format="onnx")

make sure to rename the output onnx model and run the functions separately

[inisis]

@Kayzwer Hi, CI failed, but I think it's not a problem with onnxslim, and how can I compare the performance with/without onnxslim, and one thing to mention, onnxslim can be used in arm devices.

from ultralytics import YOLO

model = YOLO("yolov8n.pt")
model.export(format="onnx", slim=True) # slimmed

model.export(format="onnx")

make sure to rename the output onnx model and run the functions separately

@Kayzwer thank you! But I think I have seen the benchmark performance before, and it's generated by CI.

[inisis]

@inisis pull request for the integration, thanks for the tool :)

@glenn-jocher can you please review it

[glenn-jocher]

@inisis Thanks for the heads up! The PR is on Glenn's radar, and he'll review it as soon as possible. Thanks for your contribution! 🚀

[Kayzwer]

@inisis Thanks for the heads up! The PR is on Glenn's radar, and he'll review it as soon as possible. Thanks for your contribution! 🚀

thanks LLM