def __init__(self):
FLAGS.max_epoch = 15
FLAGS.max_max_epoch = 30
FLAGS.max_num_threads = 20 #for 4G gpu should be ok?
FLAGS.dataset = 'one_oa'
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
#Overwrite the FLAGS from the user according to the flags
#defined in this configuration
FLAGS.feature_type = 'app' #flow or app or both
FLAGS.network = 'logits' #or logits_noisy
FLAGS.learning_rate = 0.01
training_objects, validate_objects, test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.max_epoch = 15 #50
FLAGS.max_max_epoch = 30
FLAGS.dataset = '../../../emerald/tmp/remote_images/continuous_expert'
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
FLAGS.window_size = 300
FLAGS.batch_size_fnn = 2
FLAGS.max_num_threads = 5
def __init__(self):
FLAGS.max_epoch = 15
FLAGS.max_max_epoch = 30
FLAGS.dataset = '../../../emerald/tmp/remote_images/wall_expert'
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
FLAGS.batchwise_learning = True
FLAGS.finetune = True
FLAGS.init_model_dir = '/esat/qayd/kkelchte/tensorflow/lstm_logs/cwall_batchwise'
def __init__(self):
FLAGS.max_epoch = 15 # let the learning rate decay faster
FLAGS.max_max_epoch = 30
FLAGS.dataset = 'selected'
if FLAGS.fc_only:
FLAGS.hidden_size = 400
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.cut_end = 20 #100
FLAGS.max_epoch = 50 #leave 50 instead of 10 in order to make sure it convergers to a minimum
FLAGS.max_max_epoch = 100
FLAGS.batch_size_fnn = 100
FLAGS.fc_only = True
FLAGS.hidden_size = 400
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.max_epoch = 100
FLAGS.max_max_epoch = 200
FLAGS.window_size = 20
FLAGS.batch_size_fnn = 32
FLAGS.network = 'no_cnn_depth'
FLAGS.dataset = 'sequential_oa_depth' #'sequential_oa'
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.max_epoch = 50 #leave 50 instead of 10 in order to make sure it convergers to a minimum
FLAGS.max_max_epoch = 100
FLAGS.batch_size_fnn = 100
FLAGS.fc_only = True
FLAGS.hidden_size = 400
#FLAGS.dataset='inc_fc_dagger' #'sequential_oa_rec'
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.max_epoch = 50 #30 #15 #30
FLAGS.max_max_epoch = 100 #60 #30 #60
FLAGS.max_num_threads = 20 #for 4G gpu should be ok?
#FLAGS.dataset='inc_lstm_dagger' #'sequential_oa_rec'
#FLAGS.window_size = 20
#FLAGS.batch_size_fnn = 32
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.max_epoch = 100
FLAGS.max_max_epoch = 200
FLAGS.batch_size_fnn = 100
FLAGS.network = 'no_cnn_depth'
FLAGS.fc_only = True
FLAGS.hidden_size = 400
FLAGS.dataset = 'sequential_oa_depth' #sequential_oa
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.sliding_window = True
FLAGS.max_epoch = 50
FLAGS.max_max_epoch = 100
#FLAGS.dataset='sequential_oa'
FLAGS.data_type = 'grouped'
FLAGS.window_size = 20
#FLAGS.batch_size_fnn = 32
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.max_grad_norm = 1
#FLAGS.cut_end=100
FLAGS.max_epoch = 50
FLAGS.max_max_epoch = 100
FLAGS.batch_size_fnn = 100
FLAGS.fc_only = True
FLAGS.hidden_size = 400
#FLAGS.dataset='sequential_oa'
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.max_epoch = 15 #50
FLAGS.max_max_epoch = 100
#FLAGS.hidden_size = 10 #dimensionality of cell state and output
#FLAGS.max_epoch = 2
#FLAGS.max_max_epoch = 5
FLAGS.continuous = False
FLAGS.dataset = '../../../emerald/tmp/remote_images/discrete_expert_2'
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
FLAGS.finetune = True
FLAGS.init_model_dir = '/esat/qayd/kkelchte/tensorflow/lstm_logs/dis_wsize_300'
def __init__(self):
FLAGS.max_epoch = 50
FLAGS.max_max_epoch = 100
FLAGS.dataset = 'sequential_oa'
FLAGS.data_type = 'grouped'
FLAGS.batchwise_learning = True
FLAGS.sample = 5
FLAGS.preloading = False
FLAGS.window_size = 20
FLAGS.batch_size_fnn = 32
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
FLAGS.max_num_windows = 500
FLAGS.max_epoch = 100
FLAGS.max_max_epoch = 200
FLAGS.window_size = 20
FLAGS.batch_size_fnn = 32
FLAGS.preloading = False
FLAGS.conv_layers = True
FLAGS.network = 'no_cnn'
#FLAGS.normalized = True
FLAGS.dataset = 'seq_oa_huge' #sequential_oa_rec #'sequential_oa'
self.training_objects, self.validate_objects, self.test_objects = pilot_data.get_objects(
)
def __init__(self):
training_objects, validate_objects, test_objects = pilot_data.get_objects(
)