https://github.com/jafagervik/torchtemplate

Template for working with pytorch
https://github.com/jafagervik/torchtemplate

Last synced: 3 months ago
JSON representation

Template for working with pytorch

• Host: GitHub
• URL: https://github.com/jafagervik/torchtemplate
• Owner: Jafagervik
• License: mit
• Created: 2023-05-26T00:00:14.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2023-05-26T01:44:49.000Z (about 2 years ago)
• Last Synced: 2025-02-09T20:47:06.431Z (5 months ago)
• Language: Python
• Size: 11.7 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE-MIT

Awesome Lists containing this project

README

        
# Pytorch project template 

![alt text](https://external-content.duckduckgo.com/iu/?u=https%3A%2F%2Fadityassrana.github.io%2Fblog%2Fimages%2Fcopied_from_nb%2Fimages%2Fmeme2.png&f=1&nofb=1&ipt=c660ae09b7a4e4e30eb266059751d9b9a52fbc0d95a3d04a63435fa7ae7d8965&ipo=images)

## Inspiration

![url](https://github.com/mrdbourke/pytorch-deep-learning/tree/main/going_modular/going_modular)

## Motivation

I don't want to write the same boring dataloaders all over again

I just want to be able to transform cool ideas in to code. 

Will probably use this in the course "Pattern Recognition" at Waseda 

## Structure

### model.py

Contains *ONLY* your model

### train.py

Entry point. here we just mix everything and run it all 

### detect.py

TODO: must be implemented before anything else 

### config.py

Contains all hyperparameters

And yes, we're using a python file and not yaml 

### data 

Folder to store your train, test and val images

### utils/helpers.py

Annoying code you need to get up and running

### utils/visuals.py

Don't ask me, I am not a data scientist or analyst

Some cool plotting goes into here 

### datasetup.py

Set up and install your data inside here

### eval.py

Run your own data here to only test

##  Requirements

1. You should have torch torchvision matplotlib, numpy and tqdm installed

2. TK, qt or some way of showing matplotlib

3. It's better to run with NVIDIA!