Train CIFAR10,CIFAR100 with Pytorch-lightning

Measure CNN,VGG,Resnet,WideResnet models' accuracy on dataset CIFAR10,CIFAR100 using pytorch-lightning.

Requirements

• setup/requirements.txt

torch 1.5.1
torchvision 0.6.1
pytorch-lightning 0.9.0rc5
tqdm
argparse
pytablewriter
seaborn
enum34
scipy
cffi
sklearn

• install requirements using pip

pip3 install -r setup/requirements.txt

How to run

After you have cloned the repository, you can train each models with datasets cifar10, cifar100. Trainable models are VGG, Resnet, WideResnet, Densenet-BC, Densenet.

python train.py

Implementation Details

• CIFAR10

epoch	learning rate	weight decay	Optimizer	Momentum	Nesterov
0 ~ 20	0.1	0.0005	SGD	0.9	False
21 ~ 40	0.01	0.0005	SGD	0.9	False
41 ~ 60	0.001	0.0005	SGD	0.9	False

• CIFAR100

epoch	learning rate	weight decay	Optimizer	Momentum	Nesterov
0 ~ 60	0.1	0.0005	SGD	0.9	False
61 ~ 120	0.01	0.0005	SGD	0.9	False
121 ~ 180	0.001	0.0005	SGD	0.9	False

Accuracy

Below is the result of the test set accuracy for CIFAR-10, CIFAR-100 dataset training

Accuracy of models trained on CIFAR10

network	dropout	preprocess	parameters	accuracy(%)
VGG16	0	meanstd	14M	91.09
Resnet-50	0	meanstd	23M	92.11
WideResnet 28x10	0.3	meanstd	36M	93.61
Densenet-BC	0	meanstd	769K	92.85
Densenet	0	meanstd	769K	93.06

Accuracy of models trained on CIFAR100

network	dropout	preprocess	parameters	accuracy(%)
VGG16	0	meanstd	14M	72.79
Resnet-50	0	meanstd	23M	75.80
WideResnet 28x20	0.3	meanstd	145M	75.46
Densenet-BC	0	meanstd	800K	72.23
Densenet	0	meanstd	800K	75.58