def setup_arch_flag(cls): flags.DEFINE_float('size_eps', default=0.25, help='eps for inequality method') flags.DEFINE_float('size_eps_size', default=0.2, help='eps_size for interval size band') flags.DEFINE_float('size_p_p', default=10, help='penalty for p_positive') flags.DEFINE_float('size_p_n', default=10, help='penalty for p_negative')
def setup_arch_flags(cls): super().setup_arch_flags() flags.DEFINE_integer('max_epoch', default=200, help='number of max_epoch') flags.DEFINE_multi_integer('milestones', default=[ 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 140, 160 ], help='miletones for lr_decay') flags.DEFINE_float('gamma', default=0.85, help='gamma for lr_decay') flags.DEFINE_string('device', default='cpu', help='cpu or cuda?') flags.DEFINE_integer('printfreq', default=5, help='how many output for an epoch') flags.DEFINE_integer('num_admm_innerloop', default=2, help='how many output for an epoch') flags.DEFINE_integer('num_workers', default=1, help='how many output for an epoch') flags.DEFINE_integer('batch_size', default=1, help='how many output for an epoch') flags.DEFINE_boolean('vis_during_training', default=False, help='matplotlib plot image during training') flags.DEFINE_boolean('group', default=False, help='group patient to perform 3D dice')
def setup_arch_flags(cls): """ Setup the arch_hparams """ flags.DEFINE_float('weight_decay', default=0, help='decay of learning rate schedule') flags.DEFINE_float('lr', default=0.001, help='learning rate') flags.DEFINE_boolean('amsgrad', default=False, help='amsgrad') flags.DEFINE_integer('optim_inner_loop_num', default=5, help='optim_inner_loop_num') flags.DEFINE_string('arch', default='enet', help='arch_name') flags.DEFINE_integer('num_classes', default=2, help='num of classes') flags.DEFINE_string('method', default='admm_gc_size', help='arch_name') flags.DEFINE_boolean('ignore_negative', default=False, help='ignore negative examples in the training')
def setup_arch_flag(cls): flags.DEFINE_float('reg_eps', default=0.25, help='eps for inequality method') flags.DEFINE_float('reg_p_p', default=1, help='penalty for p_positive') flags.DEFINE_float('reg_p_n', default=1, help='penalty for p_negative') flags.DEFINE_float('reg_lamda', default=1, help='lamda for unary term and pairwise term') flags.DEFINE_float('reg_sigma', default=0.02, help='smoothness term for pairwise term') flags.DEFINE_integer('reg_dilation_level', default=10, help='dilation for the weak mask') flags.DEFINE_integer('reg_kernelsize', default=5, help='dilation for the weak mask')
def setup_arch_flags(cls): super().setup_arch_flags() flags.DEFINE_float( 'lamda', default=0.01, help='balance between the unary and the neighor term') flags.DEFINE_float('sigma', default=0.001, help='Smooth the neigh term') flags.DEFINE_integer('kernelsize', default=5, help='kernelsize of the gc') flags.DEFINE_integer( 'dilation_level', default=6, help='iterations to execute the dilation operation') flags.DEFINE_integer('stop_dilation_epoch', default=70, help='stop dilation operation at this epoch')
def setup_arch_flags(cls): super().setup_arch_flags() flags.DEFINE_boolean( 'individual_size_constraint', default=True, help='Individual size constraint for each input image') flags.DEFINE_float('eps', default=0.001, help='Individual size eps') flags.DEFINE_integer( 'global_upbound', default=2000, help='global upper bound if individual_size_constraint is False') flags.DEFINE_integer( 'global_lowbound', default=20, help='global lower bound if individual_size_constraint is False') SizeConstraint.setup_arch_flag()