def __init__(self):
self.learning_rate_decay_steps = int(
parse_dict.get("learning_rate_decay_steps", "10000000"))
self.learning_rate_decay_rate = float(
parse_dict.get("learning_rate_decay_steps", "0.9"))
self.dropout_keep_deep = [1, 1, 1, 1, 1]
self.hidden_units = [
int(i) for i in parse_dict.get("hidden_units", "").split(",")
]
self.max_data_size = int(parse_dict.get("max_data_size", "8000000"))
self.learning_rate = float(parse_dict.get("learning_rate", "0.001"))
self.l2_reg = float(parse_dict.get("l2_reg", "0.00001"))
self.batch_size = 1024
self.hidden_units = [512, 256, 128]
self.alg_name = parse_dict.get("alg_name", "dnn_tensorboard")
self.action_type = parse_dict.get("action_type", "train")
self.epochs = int(parse_dict.get("epochs", "4"))
self.batch_size = int(parse_dict.get("batch_size", ""))
self.embedding_size = int(parse_dict.get("embedding_size", "16"))
self.cate_feats_size = int(parse_dict.get("cate_feats_size", "10000"))
self.num_batch_size = int(parse_dict.get("num_batch_size", "100"))
self.feat_conf_path = parse_dict.get("feat_conf_path", None)
self.train_path = parse_dict.get("train_path", "")
self.predict_path = parse_dict.get("predict_path", "")
self.model_pb = parse_dict.get("model_pb", "")
self.save_model_checkpoint = parse_dict.get("save_model_checkpoint",
"")
self.model_restore = parse_dict.get("model_restore", "0")
self.restore_model_checkpoint = parse_dict.get(
"restore_model_checkpoint", "")
self.cont_field_size, self.vector_feats_size, self.cate_field_size, self.multi_feats_size, \
self.multi_field_size, self.multi_feats_range = my_utils.feat_size(self.feat_conf_path, self.alg_name)
def __init__(self):
self.alg_name = sys.argv[1]
self.action_type = sys.argv[2]
self.epochs = int(sys.argv[3])
self.embedding_size = int(sys.argv[4])
self.cate_feats_size = int(sys.argv[5])
self.num_batch_size = int(sys.argv[6])
self.feat_conf_path = sys.argv[7]
self.train_path = sys.argv[8]
self.predict_path = sys.argv[9]
self.model_pb = sys.argv[10]
self.save_model_checkpoint = sys.argv[11]
self.model_restore = int(sys.argv[12])
self.restore_model_checkpoint = sys.argv[13]
# self.summary_log = sys.argv[14]
self.learning_rate = 0.001
self.hidden_units = [512, 256, 128]
self.dropout_keep_deep = [1, 1, 1, 1, 1]
self.learning_rate_decay_steps = 10000000
self.learning_rate_decay_rate = 0.9
self.l2_reg = 0.00001
self.batch_size = 1024
self.cont_field_size, self.vector_feats_size, self.cate_field_size, self.multi_feats_size, \
self.multi_field_size, self.multi_feats_range = my_utils.feat_size(self.feat_conf_path, self.alg_name)
def __init__(self):
self.learning_rate_decay_steps = int(
parse_dict.get("learning_rate_decay_steps", "10000000"))
self.learning_rate_decay_rate = float(
parse_dict.get("learning_rate_decay_steps", "0.9"))
# self.hidden_units = [int(i) for i in parse_dict.get("hidden_units", "").split(",")]
self.max_data_size = int(parse_dict.get("max_data_size", "8000000"))
self.learning_rate = float(parse_dict.get("learning_rate", "0.001"))
self.l2_reg = float(parse_dict.get("l2_reg", "0.00001"))
self.batch_size = int(parse_dict.get("batch_size", "1024"))
self.embedding_size = int(parse_dict.get("embedding_size", "64"))
self.cate_feats_size = int(parse_dict.get("cate_feats_size", "200000"))
self.num_batch_size = int(parse_dict.get("num_batch_size", "100"))
self.autoint_layer_num = int(parse_dict.get("autoint_layer_num", "2"))
self.autoint_emb_size = int(parse_dict.get("autoint_emb_size", "32"))
self.autoint_head_num = int(parse_dict.get("autoint_head_num", "2"))
self.autoint_use_res = int(parse_dict.get("autoint_use_res", "1"))
self.experts_units = int(parse_dict.get("experts_units ", "1024"))
self.experts_num = int(parse_dict.get("experts_num ", "8"))
self.label1_weight = float(parse_dict.get("label1_weight", "0.5"))
self.label2_weight = float(parse_dict.get("label2_weight", "0.5"))
self.action_type = parse_dict.get("action_type", "train")
self.epochs = int(parse_dict.get("epochs", "2"))
self.alg_name = parse_dict.get("alg_name", "mmoe")
self.feat_conf_path = parse_dict.get("feat_conf_path",
"files/conf/esmm")
self.train_path = parse_dict.get("train_path",
"files/data/esmm/train/")
self.predict_path = parse_dict.get("predict_path",
"files/data/esmm/pred/")
self.model_pb = parse_dict.get("model_pb",
"files/model_save_pb/" + self.alg_name)
# self.model_pb = parse_dict.get("model_pb", "D:\\PycharmProjects\\deep_learing_estimator\\files\\model_save_pb\\" + self.alg_name)
self.model_dir = parse_dict.get(
"model_dir", "files/model_save_dir/" + self.alg_name)
self.hidden_units = [1024, 512, 256]
self.dropout_keep_deep = [1, 1, 1, 1, 1]
self.dropout_keep_fm = [1, 1, 1, 1, 1]
self.is_GPU = 1
self.num_cpu = 10
self.log_step_count_steps = 500
self.save_checkpoints_steps = 500
self.save_summary_steps = 500
self.keep_checkpoint_max = 2
self.rate = 1.0
self.cont_field_size, self.vector_feats_size, self.cate_field_size, self.multi_feats_size, \
self.multi_feats_range, self.attention_feats_size, self.attention_range = my_utils.feat_size(self.feat_conf_path,
self.alg_name)
def __init__(self):
self.dt = sys.argv[1]
self.alg_name = sys.argv[2]
self.mode = sys.argv[3]
self.epochs = int(sys.argv[4])
self.embedding_size = int(sys.argv[5])
self.cate_feats_size = int(sys.argv[6])
self.feat_conf_path = sys.argv[7]
self.train_path = sys.argv[8] + self.dt + "/"
self.predict_path = sys.argv[9] + self.dt + "/"
self.model_pb = sys.argv[10] + self.dt + "/"
self.model_dir = sys.argv[11]
self.is_GPU = int(sys.argv[12])
hidden_units_str = sys.argv[13]
self.hidden_units = list() # self.hidden_units = [1024, 512, 256]
self.learning_rate = 0.001
self.dropout_keep_deep = [1, 1, 1, 1, 1]
self.dropout_keep_fm = [1, 1, 1, 1, 1]
self.learning_rate_decay_steps = 10000000
self.learning_rate_decay_rate = 0.9
self.l2_reg = 0.00001
self.batch_size = 1024
self.num_cpu = 20
self.log_step_count_steps = 500
self.save_checkpoints_steps = 100
self.save_summary_steps = 500
self.keep_checkpoint_max = 3
self.cont_field_size, self.vector_feats_size, self.cate_field_size, self.multi_feats_size, \
self.multi_feats_range, self.attention_feats_size, self.attention_range = my_utils.feat_size(self.feat_conf_path, self.alg_name)
hidden_arr = hidden_units_str.split(",")
for i in range(len(hidden_arr)):
self.hidden_units.append(int(hidden_arr[i]))