def parse_model_args(parser):
parser.add_argument(
'--stage',
type=int,
default=2,
help='Stage of training: 1-KG_pretrain, 2-recommendation.')
parser.add_argument(
'--base_method',
type=str,
default='BPR',
help='Basic method to generate recommendations: BPR, GMF')
parser.add_argument('--emb_size',
type=int,
default=64,
help='Size of embedding vectors.')
parser.add_argument('--time_scalar',
type=int,
default=60 * 60 * 24 * 100,
help='Time scalar for time intervals.')
parser.add_argument(
'--category_col',
type=str,
default='i_category',
help='The name of category column in item_meta.csv.')
parser.add_argument(
'--lr_scale',
type=float,
default=0.1,
help='Scale the lr for parameters in pre-trained KG model.')
parser.add_argument('--margin',
type=float,
default=1,
help='Margin in hinge loss.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size',
type=int,
default=64,
help='Size of embedding vectors.')
parser.add_argument('--gamma',
type=float,
default=1,
help='Coefficient of the contrastive loss.')
parser.add_argument(
'--beta_a',
type=int,
default=3,
help='Parameter of the beta distribution for sampling.')
parser.add_argument(
'--beta_b',
type=int,
default=3,
help='Parameter of the beta distribution for sampling.')
parser.add_argument(
'--ctc_temp',
type=float,
default=1,
help='Temperature in context-target contrastive loss.')
parser.add_argument(
'--ccc_temp',
type=float,
default=0.2,
help='Temperature in context-context contrastive loss.')
parser.add_argument(
'--encoder',
type=str,
default='BERT4Rec',
help='Choose a sequence encoder: GRU4Rec, Caser, BERT4Rec.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size',
type=int,
default=64,
help='Size of embedding vectors.')
parser.add_argument('--mip_weight',
type=float,
default=0.2,
help='Coefficient of the MIP loss.')
parser.add_argument('--sp_weight',
type=float,
default=0.5,
help='Coefficient of the SP loss.')
parser.add_argument(
'--mask_ratio',
type=float,
default=0.2,
help='Proportion of masked positions in the sequence.')
parser.add_argument(
'--stage',
type=int,
default=1,
help=
'Stage of training: 1-pretrain, 2-finetune, default-from_scratch.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size',
type=int,
default=64,
help='Size of embedding vectors.')
parser.add_argument('--attn_size',
type=int,
default=8,
help='Size of attention vectors.')
parser.add_argument('--K',
type=int,
default=2,
help='Number of hidden intent.')
parser.add_argument('--add_pos',
type=int,
default=1,
help='Whether add position embedding.')
parser.add_argument('--temp',
type=float,
default=1,
help='Temperature in knowledge distillation loss.')
parser.add_argument('--n_layers',
type=int,
default=1,
help='Number of the projection layer.')
parser.add_argument(
'--stage',
type=int,
default=3,
help=
'Stage of training: 1-pretrain_extractor, 2-pretrain_predictor, 3-joint_finetune.'
)
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size', type=int, default=64,
help='Size of embedding vectors.')
parser.add_argument('--hidden_size', type=int, default=100,
help='Size of hidden vectors in GRU.')
parser.add_argument('--attention_size', type=int, default=50,
help='Size of attention hidden space.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size',
type=int,
default=64,
help='Size of embedding vectors.')
parser.add_argument('--t_scalar',
type=int,
default=60,
help='Time interval scalar.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size', type=int, default=64,
help='Size of embedding vectors.')
parser.add_argument('--attn_size', type=int, default=8,
help='Size of attention vectors.')
parser.add_argument('--K', type=int, default=2,
help='Number of hidden intent.')
parser.add_argument('--add_pos', type=int, default=1,
help='Whether add position embedding.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size',
type=int,
default=64,
help='Size of embedding vectors.')
parser.add_argument('--num_layers',
type=int,
default=1,
help='Number of self-attention layers.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size',
type=int,
default=64,
help='Size of embedding vectors.')
parser.add_argument('--temp',
type=float,
default=0.2,
help='Temperature in contrastive loss.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size',
type=int,
default=64,
help='Size of embedding vectors.')
parser.add_argument('--num_horizon',
type=int,
default=16,
help='Number of horizon convolution kernels.')
parser.add_argument('--num_vertical',
type=int,
default=8,
help='Number of vertical convolution kernels.')
parser.add_argument('--L',
type=int,
default=4,
help='Union window size.')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size', type=int, default=64,
help='Size of embedding vectors.')
parser.add_argument('--neg_head_p', type=float, default=0.5,
help='The probability of sampling negative head entity.')
parser.add_argument('--num_layers', type=int, default=1,
help='Number of self-attention layers.')
parser.add_argument('--num_heads', type=int, default=1,
help='Number of attention heads.')
parser.add_argument('--gamma', type=float, default=-1,
help='Coefficient of KG loss (-1 for auto-determine).')
parser.add_argument('--attention_size', type=int, default=10,
help='Size of attention hidden space.')
parser.add_argument('--pooling', type=str, default='average',
help='Method of pooling relational history embeddings: average, max, attention')
parser.add_argument('--include_val', type=int, default=1,
help='Whether include relation value in the relation representation')
return SequentialModel.parse_model_args(parser)
def parse_model_args(parser):
parser.add_argument('--emb_size', type=int, default=64,
help='Size of embedding vectors.')
return SequentialModel.parse_model_args(parser)