MNIST Classification

This project fine-tune torchvision pretrained weights to classify MNIST images and try to run Pytorch model with Rust

Environments

• Python 3.8.10
• torch 2.0.1
• torchvision 0.15.2

Install requirements

pip install -U pip
pip install -r requirements.txt

Data

You can download MNIST dataset from here, use torch dataset or simply run this script

bash scripts/download_data.sh

Config

Modify config in ./configs/config.py

Train

Simply run

python tools/pytorch/train.py

Experiment Results

Some experiment results

Model	Accuracy	Confusion Matrix	Pretrained	Model size
Resnet18	99.64%		Model	44.71MB

You can download weight file above

Infer

You can infer with

python tools/pytorch/infer.py

Inference on other format

Rust

Refer here

Some inference results

You can try on your own 😉

Reference

• tch-rs