stable-continual-learning
stable-continual-learning copied to clipboard
Change the hyper-parameters in split-CIFAR
In the scripts replicate_experiment_2.sh
here, to run the split-CIFAR experiments, the command is
python -m stable_sgd.main --dataset cifar100 --tasks 20 --epochs-per-task 1 --lr 0.1 --gamma 0.8 --hiddens 256 --batch-size 10 --dropout 0.5 --seed 1234
which only provides results around 2x% average accuracy
I found in a closed issue here that the setting should be
python -m stable_sgd.main --dataset cifar100 --tasks 20 --epochs-per-task 1 --lr 0.15 --gamma 0.85 --batch-size 10 --dropout 0.1 --seed 1234
Maybe the scripts need some updates @imirzadeh?