# "C-tanh",
"bounce-states",
# "bounce-images",
]
exp_id = "C-May16"
exp_mode = "meta"
# exp_mode = "finetune"
# exp_mode = "oracle"
is_VAE = False
is_uncertainty_net = False
is_regulated_net = False
is_load_data = False
VAE_beta = 0.2
task_id_list = get_args(task_id_list, 3, type = "tuple")
if task_id_list[0] in ["C-sin", "C-tanh"]:
statistics_output_neurons = 2 if task_id_list[0] == "C-sin" else 4
z_size = 2 if task_id_list[0] == "C-sin" else 4
num_shots = 10
input_size = 1
output_size = 1
reg_amp = 1e-6
forward_steps = [1]
is_time_series = False
elif task_id_list[0] in ["bounce-states", "bounce-states2"]:
statistics_output_neurons = 8
num_shots = 100
z_size = 8
input_size = 6
output_size = 2
batch_size_task = min(50, num_train_tasks) num_backwards = 1 num_iter = 10000 pre_pooling_neurons = 200 num_context_neurons = 0 statistics_pooling = "max" main_hidden_neurons = (40, 40) patience = 200 reg_amp = 1e-6 activation_gen = "leakyRelu" activation_model = "leakyRelu" optim_mode = "indi" loss_core = "huber" array_id = "new" exp_id = get_args(exp_id, 1) exp_mode = get_args(exp_mode, 2) task_id_list = get_args(task_id_list, 3, type = "tuple") statistics_output_neurons = get_args(statistics_output_neurons, 4, type = "int") is_VAE = get_args(is_VAE, 5, type = "bool") VAE_beta = get_args(VAE_beta, 6, type = "float") lr = get_args(lr, 7, type = "float") batch_size_task = get_args(batch_size_task, 8, type = "int") pre_pooling_neurons = get_args(pre_pooling_neurons, 9, type = "int") num_context_neurons = get_args(num_context_neurons, 10, type = "int") statistics_pooling = get_args(statistics_pooling, 11) main_hidden_neurons = get_args(main_hidden_neurons, 12, "tuple") reg_amp = get_args(reg_amp, 13, type = "float") activation_gen = get_args(activation_gen, 14) activation_model = get_args(activation_model, 15) optim_mode = get_args(optim_mode, 16)
num_train_tasks = 100 num_test_tasks = 100 batch_size_task = 100 num_backwards = 1 num_iter = 20000 pre_pooling_neurons = 100 num_context_neurons = 0 statistics_pooling = "max" patience = 400 reg_amp = 1e-6 activation_gen = "leakyRelu" activation_model = "leakyRelu" optim_mode = "sum" array_id = "0" exp_id = get_args(exp_id, 1) task_id_list = get_args(task_id_list, 2, type="tuple") statistics_output_neurons = get_args(statistics_output_neurons, 3, type="int") is_VAE = get_args(is_VAE, 4, type="bool") VAE_beta = get_args(VAE_beta, 5, type="float") lr = get_args(lr, 6, type="float") batch_size_task = get_args(batch_size_task, 7, type="int") pre_pooling_neurons = get_args(pre_pooling_neurons, 8, type="int") num_context_neurons = get_args(num_context_neurons, 9, type="int") statistics_pooling = get_args(statistics_pooling, 10) reg_amp = get_args(reg_amp, 11, type="float") activation_gen = get_args(activation_gen, 12) activation_model = get_args(activation_model, 13) optim_mode = get_args(optim_mode, 14) is_uncertainty_net = get_args(is_uncertainty_net, 15, "bool") array_id = get_args(array_id, 16)