pyramidnet_randwidth.yml

# data
dataset: cifar100
workers: 4

# info
num_classes: 100
image_size: 32
topk: [1, 5]
start_epoch: 0
epochs: 300
print_freq: 100
batch_size: 64

# optimizer
optimizer: sgd
momentum: 0.9
weight_decay: 0.0001
nesterov: True

# lr
lr: 0.25
lr_scheduler: multistep
multistep_lr_milestones: [150, 225]

# model
model: pyramidnet
depth: 200
alpha: 240
max_width: 1.0
min_width: 0.8
num_subnet: 3
resos: [32, 28, 24]
reset_parameters: True
resume: ''
log_dir: logs/pyramidnet200_randwidth/
test_only: False
pretrained: ''