tinyim_fedavgm_feddecorr.sh

device_id=0
MODEL=mobilenetv2
ALPHA=0.05
PARTITION=noniid
WD=1e-4

DECORR_COEF=0.1

CUDA_VISIBLE_DEVICES=$device_id python3 main.py \
    --dataset=tinyimagenet \
    --model=$MODEL \
    --approach=fedoptim \
    --auto_aug \
    --lr=0.01 \
    --weight_decay=$WD \
    --epochs=10 \
    --n_comm_round=50 \
    --print_interval=25 \
    --n_parties=10 \
    --partition=$PARTITION \
    --alpha=$ALPHA \
    --logdir='./logs/' \
    --datadir='./data/tiny-imagenet-200/' \
    --ckptdir='./models/' \
    --server_optimizer=gd \
    --server_momentum=0.5 \
    --feddecorr \
    --feddecorr_coef=$DECORR_COEF