Smart-Pruner

Overview

Smart-Pruner is a Pytorch based module designed to apply various pruning techniques across multiple model-dataset architectures. It simplifies the evaluation process by providing insights into accuracy, energy consumption, and speedup for different pruning techniques.

Features

• Apply multiple pruning techniques to diverse model-dataset configurations.
• Obtain insights on accuracy, energy consumption, and speedup.
• Easy integration of new pruning methods.

Installation

1. Clone the Smart-Pruner repository:

   git clone https://github.com/prabhas2002/Smart-Pruner.git

2 Install the required dependencies:

pip install -r requirements.txt

Usage

Check the ipynb files for example usage.

Pruning Methods

• The pruning methods supported by Smart-Pruner, such as:
  • Global Pruning
  • Random Unstructured Pruning
  • L1-Norm Based Filter pruning
  • Ln Structured pruning
  • Pruning with 2:4 Sparsity (check research paper)
  • Decay Pruning
  • Thinet (check research paper)

Architectures

• The model-dataset architectures compatible with Smart-Pruner, for example:
  • ResNet-50 with CIFAR-10
  • AlexNet with CIFAR10
  • LeNet on MNIST
  • VggNet on CIFAR10

Incorporating New Pruning Methods

• Smart-Pruner allows easy integration of new pruning methods. Follow these steps:
  1. Implement your new pruning method in the /pruning folder.
  2. Add the new pruning file name in init file of it.
  3. Import it in ipynb file.

Contributing

We welcome contributions to Smart-Pruner! If you have suggestions, bug reports, or want to add new features, please fork the repository and submit a pull request.