MNIST

A toy based on libtorch API.

1. Prerequisite

This project is managed by Pixi. Before moving on, please ensure it is installed already.

Target platform: linux-64

Target device: cpu

2. Getting Started

Get Data

Get dataset and do preprocess:

pixi install -e dataset
pixi r get-dataset
pixi r process-dataset

Configure and Build

Configure and build the project:

pixi install -e default
pixi r configure [release|debug]
pixi r build

The task configure defaults to debug. For release configure and build, type:

pixi r configure release

Train and Test

Train the network:

pixi r train <path-to-train-config-file>

For example:

pixi r train configs/train.toml

The <path-to-train-config-file> defaults to configs/train.toml.

Test the trained network:

pixi r test <path-to-test-config-file>

For example:

pixi r test configs/test.toml

The <path-to-test-config-file> defaults to configs/test.toml.

Example config file can be found in configs/config.example.toml.