hhchen1105/Supervised-Contrastive-Parallel-Learning

Supervised Contrastive Parallel Learning

Setup

Tested under Python 3.8.12 in Ubuntu. Install the required packages by

$ pip install -r requirements.txt

File Description

• tiny-imagenet-200.zip：A zip file of tinyImageNet dataset processed under PyTorch ImageFolder format. Download link: https://drive.google.com/file/d/1R5QMeXAL_8XYqaDiGFFFoM1IwiJ5ZcBJ/view?usp=sharing

Quick Start

Unzip tiny-imagenet-200.zip

$ unzip tiny-imagenet-200.zip

Config Settings

• data section
  • dataset: dataset for experiments. Options: cifar10, cifar100 or tinyImageNet
  • train_batch_size: batch size for training
  • test_batch_size: batch size for testing
  • augmentation: use basic augmentation like BP commonly used, or strong augmentation like contrastive learning used. Options: basic, strong
• model section
  • model: model and loss functions for experiments. Options: CNN, CNN_AL, CNN_SCPL, CNN_PredSim, VGG, VGG_AL, VGG_SCPL, VGG_PredSim, resnet, resnet_AL, resnet_SCPL, resnet_PredSim
  • epochs: number of epochs for training
  • base_lr: initial learning rate
  • end_lr: learning rate at the end of training
• Example

[data]
dataset = tinyImageNet
train_batch_size = 128
test_batch_size = 1024
augmentation = strong


[model]
model = VGG
epochs = 200
base_lr = 0.001
end_lr = 0.00001

Execute

$ python main.py

• Then, the tensorboard logger results will be saved in {work_dir}/{model}/{dataset}/tb_{i} folder where i means the i-th experiment