test_UniP

This repository aimed to provide the test models and test the pytorch pruning framework UniP.

Introduction

This repository is based on the UniP, and help users of UniP to test the pruning framework. If you could not run the UniP with your own models, you should pull a request to this repository. Let's fix your problems and make UniP better.

Requirements

torch
torchvision
UniP

Getting Started

1. Install UniP

git clone https://github.com/Nobreakfast/test_UniP.git
cd test_UniP
pip install -e .

2. Folder structure:

test_UniP
├── test_unip
│   ├── models
│   │   ├── template_model.py
│   │   ├── example_model.py
│   │   └── ...
├── tests
│   ├── test_template_model.py
│   ├── test_example_model.py
│   └── ...
├── requirements
│   ├── template_requirements.txt
│   ├── example_requirements.txt
│   └── ...
├── requirements.txt
└── README.md

3. Steps to add your model and test script:

Fork this repository to your github account.
Clone your forked repository to your local machine.

cd /path/to/your/disired/directory
git clone https://github.com/your_github_account/test_UniP.git

🔥 Important 🔥 Put your model in the models directory, you could create any nick named folder or file. You could refer to the template file test_unip/models/template_model.py
🔥 Important 🔥 Add a simple test script in the test directory, you could create any nick named file (tests/test_your_file.py). You could refer to the template file tests/test_template.py. The test script should provide two functions: test_inference() and UniP test_pruning():

import unip
from unip.core.pruner import BasePruner
from unip.utils.evaluation import cal_flops
from test_unip.models.example_model import ExampleModel

def test_inference():
    # Your inference code here
    example_input = torch.randn(1, 3, 224, 224)
    model = ExampleModel()
    out = model(example_input)
def test_pruning():
    # Your pruning code here
    example_input = torch.randn(1, 3, 224, 224)
    model = ExampleModel()
    pruner = BasePruner(model, example_input, "RandomRatio")
    pruner.algorithm.run(0.8)
    pruner.prune()
    cal_flops(model, example_input, "cpu")

Make sure the test_inference() could run successfully. If your model need some external library, you should add the requirements/example_requirements.txt file to provide the requirements. You could refer to the template file requirements/template_requirements.txt.

Commit your changes and push to your forked repository.

git add .
git commit -m "add new model"
git push

Open your forked respository on github, and pull a request to this repository.
Wait for the test result. If the test_inference() function could run successfully, we will accept your request. If not, we will leave a comment on your pull request and you should fix the problem and pull a request again.
Open the UniP Issue Page and create a new issue. Describe your problem and provide the link of your PR ID and Link. We will fix your problem as soon as possible.
Once your problem is fixed, we will close your issue and your tests/test_your_file.py will be changed to tests/your_file_PRID.py. If you still facing problems, you could rechange the file name, pull a request and reopen the issues.

Changelogs

2023-09-03: Optimize the test process.