SimCLR

A PyTorch implementation of SimCLR based on ICML 2020 paper A Simple Framework for Contrastive Learning of Visual Representations.

Requirements

• Anaconda
• PyTorch

conda install pytorch torchvision cudatoolkit=10.0 -c pytorch

• thop

pip install thop

Dataset

CIFAR10 dataset is used in this repo, the dataset will be downloaded into data directory by PyTorch automatically.

Usage

Train SimCLR

python main.py --batch_size 1024 --epochs 1000 optional arguments:
--feature_dim Feature dim for latent vector [default value is 128]
--temperature Temperature used in softmax [default value is 0.5]
--k Top k most similar images used to predict the label [default value is 200]
--batch_size Number of images in each mini-batch [default value is 512]
--epochs Number of sweeps over the dataset to train [default value is 500]

Linear Evaluation

python linear.py --batch_size 1024 --epochs 200 optional arguments:
--model_path The pretrained model path [default value is 'results/128_0.5_200_512_500_model.pth']
--batch_size Number of images in each mini-batch [default value is 512]
--epochs Number of sweeps over the dataset to train [default value is 100]

Results

There are some difference between this implementation and official implementation, the model (ResNet50) is trained on one NVIDIA TESLA V100(32G) GPU:

1. No Gaussian blur used;
2. Adam optimizer with learning rate 1e-3 is used to replace LARS optimizer;
3. No Linear learning rate scaling used;
4. No Linear Warmup and CosineLR Schedule used.

Evaluation Protocol	Params (M)	FLOPs (G)	Feature Dim	Batch Size	Epoch Num	τ	K	Top1 Acc %	Top5 Acc %	Download
KNN	24.62	1.31	128	512	500	0.5	200	89.1	99.6	model | gc5k
Linear	23.52	1.30	-	512	100	-	-	92.0	99.8	model | f7j2