config.yaml

# distributed training
nodes: 1
gpus: 1 # I recommend always assigning 1 GPU to 1 node
nr: 0 # machine nr. in node (0 -- nodes - 1)
dataparallel: 0 # Use DataParallel instead of DistributedDataParallel
workers: 8
dataset_dir: "/Users/gaurav/Desktop/thesis-work/Datasets/T-1/train-faces/all/train"

# train options
seed: 42 # sacred handles automatic seeding when passed in the config
batch_size: 128
image_size: [108, 124]
start_epoch: 0
epochs: 100
dataset: "TFIW" #"CIFAR10" # STL10
pretrain: True 

# model options
resnet: "resnet18"
projection_dim: 64 # "[...] to project the representation to a 128-dimensional latent space"

# loss options
optimizer: "Adam" # or LARS (experimental)
weight_decay: 1.0e-6 # "optimized using LARS [...] and weight decay of 10−6"
temperature: 0.5 # see appendix B.7.: Optimal temperature under different batch sizes

# reload options
model_path: "save" # set to the directory containing `checkpoint_##.tar` 
epoch_num: 100 # set to checkpoint number
reload: False

# logistic regression options
logistic_batch_size: 256
logistic_epochs: 500