MNIST Classify

Introduction

The repository contains the PyTorch implementation of image classification models for MNIST dataset, based on MLP, CNN, RNN, LSTM, and GRU.

Results

Metrics

Model	#FLOPs (M)	#Params (K)	Acc (%)
RNN	0.3	6.5	94.4
LSTM	1.3	24.1	98.5
GRU	1.0	18.3	98.9
MLP	0.5	238.0	98.7
CNN	0.6	43.8	99.3

Loss & Accuracy Curves

Requirements

pip install -r ./requirements.txt

Training & Testing

Training

python train.py --arch <ARCH> [--model_name <MODEL_NAME>] [--gpu <GPU_ID>]

Test

# Test one model
python test.py --arch <ARCH> [--model_name <MODEL_NAME>] [--gpu <GPU_ID>]

# Test all models
./run.sh