/mini-torch

Naive torch implementation for unraveling the mystery of Deep Learning.

Primary LanguagePython

Mini-Torch

This repository tries to implement a naive version of torch to explain the secrets behind the deep learning framework. It is not the same as the official torch, it is just implemented in a way that borrows the api style of the torch. The exact implementation is different.

It is all python, without any c++ or cuda code, easy for readers to understand. It also has mathematical formula derivations so that the reader can know why the code is written the way it is.