mohitkr95 / cifar10_prune_project

CIFAR-10 CNN training and optimization with PyTorch: A comprehensive guide to model pruning for efficient performance.

CIFAR-10 Model Training and Pruning

This project provides an implementation for training a convolutional neural network (CNN) on the CIFAR-10 dataset and subsequently applying model pruning to reduce the model size.

Table of Contents

• Installation
• Directory Structure
• Usage
• Results
• License

Installation

1. Clone the repository:

   git clone https://github.com/Mohitkr95/CIFAR10_Prune_Project.git
   

2. Change your current directory to the project folder:

   cd CIFAR10_Prune_Project
   

3. Install required packages (Optional: It's recommended to create a virtual environment before installing the packages):

   pip install torch torchvision
   

Directory Structure

CIFAR10_Prune_Project/
│
├── data/                   
│
├── models/                 
│   ├── model.pth
│   └── pruned_model.pth
│
├── src/                    
│   ├── __init__.py
│   ├── dataset.py          
│   ├── model.py            
│   ├── train.py            
│   └── prune.py            
│
├── utils/                  
│   └── __init__.py
│
├── main.py                 
│
└── README.md               

Usage

1. Run the main script to train and prune the model:

   python main.py
   

2. The original and pruned model weights will be saved in the models/ directory.

Results

After running the main.py script, you'll get the following results:

• Training and validation accuracy of the original model.
• Validation accuracy after model pruning with different pruning ratios.
• Saved model weights for both the original and pruned models in the models/ directory.

License

MIT © Mohitkr95