Project Overview

This repository contains three main folders: homework, paper, and project.

Folder Structure

1. homework

This folder contains the homework solutions for 12 weeks, organized into 12 separate folders (one for each week). Each folder includes solutions to the corresponding week's tasks. The naming convention for each solution is as follows:

week_task_descriptiveName

• Files: For smaller tasks.
• Folders: For larger tasks that require multiple files.

2. paper

This folder contains the paper related to the project.

3. project

The project folder contains the main project work, divided into two subfolders: final and development.

a) final

The final folder represents the final state of the project. It includes:

• dist/:

  • Contains a .whl (wheel) file that allows installation of the fast_einsum package on Windows machines that support AVX2 vector instructions.

  Installation command:

  pip install dist/fast_einsum-0.1-cp311-cp311-win_amd64.whl

  Usage

You can use fast_einsum as follows:

import numpy as np
from fast_einsum import fast_einsum

A = np.random.rand(3, 3)
B = np.random.rand(3, 3)

C = np.einsum("ab,bc->ac", A, B)
C_tilda = fast_einsum("ab,bc->ac", A, B)

print(np.allclose(C, C_tilda))

• benchmark/:

  • Contains code for benchmarking the implementation of fast_einsum.

• plot/:

  • Contains code for generating plots based on the benchmarking results for the paper.

• fast_einsum/:

  • The main implementation of the fast_einsum package.

• setup.py:

  • Used to build the wheel file for the fast_einsum package.

b) development

The development folder includes all development versions of the fast_einsum package.

• main.py:
  • Benchmarks the different development versions against each other. The results of these benchmarks are presented and discussed in the paper.

---