NOTE: THE OFFICIAL CODE IS NOW MOVED TO A NEW REPOSITORY
[Not maintained] Code implementation for "Continual Semi-Supervised Learning through Contrastive Interpolation Consistency" - Accepted at Pattern Recognition Letters 2022
To run the experiments:
- export $PYTHONPATH=<ROOT DIR OF THIS REPO>
python utils/main.py
(+ args)- argument
lpc
(labels per class) specifies how many labels are not masked (leave it empty for full supervision)
For example:
python utils/main.py --n_epochs=50 --model=ccic --dataset=seq-cifar10 --lr=0.001 --batch_size=32 --buffer_size=500 --minibatch_size=32 --alpha=0.5 --lamda=0.5 --k=3 --memory_penalty=1 --k_aug=3 --sharp_temp=0.5 --mixup_alpha=0.75