forked from shenzebang/Federated-Learning-Pytorch
-
Notifications
You must be signed in to change notification settings - Fork 0
/
mnist_ablation_eps.sh
executable file
·17 lines (16 loc) · 996 Bytes
/
mnist_ablation_eps.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
DEVICE=0
########################################################################################
for minority in 3
do
for d in 0.3
do
for eps in 0.02 0.002 0.001 0.2 0.0005
do
for formulation in "imbalance-fl-res" "imbalance-fl"
do
CUDA_VISIBLE_DEVICES=$DEVICE python run_PD_FL.py --project FedResFinal --formulation ${formulation} --run alpha_${imbalance}_${formulation} --reduce_to_ratio 0.01 --perturbation_lr 0.1 --perturbation_penalty 2.0 --dataset mnist --imbalance --n_workers_per_round 100 --use_ray --learner fed-avg --local_lr 5e-2 --n_pd_rounds 500 --loss_fn cross-entropy-loss --model mlp --n_workers 100 --n_p_steps 5 --dense_hid_dims 128-128 --no_data_augmentation --lambda_lr .1 --tolerance_epsilon $eps --use_gradient_clip --n_minority $minority --heterogeneity dir --dir_level $d
done
done
done
done
########################################################################################