PyTorch Settings Alpha

This is the Alpha pre release of pytorch settings. You will find that this is mostly test code to see how it works.

These are configurations tailored for my machine in various machine learning projects. These settings are designed to streamline my workflow and ensure consistency across different environments. When I started using Pytorch, I realized that there are settings that I might have to move back and forth from different devices, beacuse of testing on different CPU sets. This is the collection of utilities I wrote that allow me to move back and forth from machine to machine and copying my settings. I am going to build a utility to drop this into something like OneDrive or FTP. You will see some of that logic already built.

Overview

This repository houses various configuration files and scripts that I utilize to set up PyTorch environments for my projects. These configurations have aspects such as device handling, seed settings for reproducibility, and performance optimizations.

Contents

• environment.yml: Contains a Conda environment file for easily setting up PyTorch with all necessary dependencies.
• settings.py: Custom Python script to configure PyTorch settings. It covers device management, reproducibility through seed settings, and performance optimization.
• utils/: This directory holds utility scripts that enhance PyTorch functionality, making tasks like data loading and model training more efficient.

Usage

To incorporate these settings into your projects, follow these steps:

1. Clone this repository:

   git clone https://github.com/bbhorrigan/pytorchsettings.git
   cd pytorchsettings

2. Set up the Conda environment using the provided environment file:

   conda env create -f environment.yml

3. Activate the created environment. Replace <environment-name> with the name you choose for the Conda environment:

   conda activate <environment-name>

   For example, if you named your environment "pytorch_env", the command would be:

   conda activate pytorch_env