E2E_model

This directory contains code and resources for end-to-end deep learning model experiments, including asymmetric and symmetric quantization, CNN architectures, and training scripts. The project is focused on image classification tasks using datasets like CIFAR-10 and MNIST.

Directory Structure

• Asymmetric.py — Implements asymmetric quantization methods.
• Symmetric.py — Implements symmetric quantization methods.
• CNNImage.py — CNN model for image classification.
• CNNToy.py — Toy CNN model for experimentation.
• CNNToyOptim.py — Optimized version of the toy CNN.
• Conv2dPt2e.py — Custom 2D convolutional layer for quantization experiments.
• PCQ.py, PGQ.py — Quantization utilities and algorithms.
• best_model.pth — Pretrained model weights.
• data/ — Contains datasets:
  • cifar-10-python.tar.gz — CIFAR-10 dataset archive.
  • cifar-10-batches-py/ — Extracted CIFAR-10 batches.
  • MNIST/ — MNIST dataset files.
• README.md — This file.

Requirements

• Python 3.8+
• PyTorch
• NumPy

(Install dependencies using pip install torch numpy.)

Usage

1. Place datasets in the data/ directory as shown above.

2. Use the provided scripts to train or evaluate models. For example:

   python CNNToy.py

   or

   python CNNImage.py

3. Pretrained weights are available in best_model.pth.

Notes

• Modify the scripts as needed for your experiments.
• See individual script files for more details on their usage and parameters.