config.py

import os
import torch

gpu = 0
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'  # ignore warning
os.environ["CUDA_VISIBLE_DEVICES"] = str(gpu)  # use gpu0,1

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
#device = 'cpu

ssl_task = True  
#ssl_task = False  

dataset = 'svhn'
random_state = 111
num_channels = 3
image_shape = 32

#########################################################################################
# # SSL parameters 

pretext_task = 'label_augmentation'

lr = 0.0001
batch_size = 64
w_d = 1e-5
momentum = 0.9
# w_d = 0

regularization = 'l2_norm'

if regularization == 'l2_norm':
    alpha = 10 #10 #0.01    #0, 2
    beta = 18 #18 # 0.1  #0

num_epochs  = 50
num_tasks   = 3   # This task referes to number attributes for multi-task network
num_classes = 2   # This refers to number of classes in each task

random_seed_list = [8, 69, 78, 121, 512]
random_seed_counter = 0
#########################################################################################
# # OWM parameters

nepochs = [25, 25, 25, 25, 25]
owm_lrs = [0.1, 0.1, 0.1, 0.1, 0.1]

num_layers = 3  # 3, 5, 8
model = 'owm_'+str(num_layers)

pretrained = True
#pretrained = False