import argparse def parse_args(): parser = argparse.ArgumentParser() # General Args parser.add_argument('--rs_type', type=str, default='General', choices=['Seq', 'LLM', 'General'], help='Seq, LLM, General') parser.add_argument('--model_name', type=str, default='SASRec', help='model name.') parser.add_argument('--dataset', nargs='?', default='yc', help='yc, ks, rr') parser.add_argument('--cuda', type=int, default=0, help='cuda device.') parser.add_argument('--test_only', action="store_true", help='Whether to test only.') parser.add_argument('--clear_checkpoints', action="store_true", help='Whether clear the earlier checkpoints.') parser.add_argument('--saveID', type=str, default='Saved', help='Specify model save path. Description of the experiment') parser.add_argument('--seed', type=int, default=101, help='Random seed.') parser.add_argument('--max_epoch', type=int, default=500, help='Number of max epochs.') parser.add_argument('--verbose', type=float, default=5, help='Interval of evaluation.') parser.add_argument('--patience', type=int, default=10, help='Early stopping point.') parser.add_argument("--mix", action="store_true", help="whether to use mixed dataset") # Model Args parser.add_argument('--batch_size', type=int, default=4096, help='Batch size.') # parser.add_argument('--batch_size', type=int, default=128, # help='Batch size.') # parser.add_argument('--lr', type=float, default=0.001, # help='Learning rate.') parser.add_argument('--lr', type=float, default=5e-4, help='Learning rate.') parser.add_argument('--hidden_size', type=int, default=64, help='Number of hidden factors, i.e., embedding size.') parser.add_argument('--weight_decay', type=float, default=1e-6, help='weight decay for optimizer.') parser.add_argument('--dropout', type=float, default=0.1, help='dropout ') parser.add_argument("--no_wandb", action="store_true", help="whether to use wandb") args, _ = parser.parse_known_args() if(args.rs_type == 'General'): parser.add_argument("--candidate", action="store_true", help="whether using the candidate set") parser.add_argument('--Ks', type = int, default= 20, help='Evaluate on Ks optimal items.') parser.add_argument('--neg_sample',type=int,default=1) parser.add_argument('--infonce', type=int, default=0, help='whether to use infonce loss or not') parser.add_argument("--train_norm", action="store_true", help="train_norm") parser.add_argument("--pred_norm", action="store_true", help="pred_norm") parser.add_argument('--n_layers', type=int, default=0, help='Number of GCN layers') parser.add_argument('--data_path', nargs='?', default='./data/General/', help='Input data path.') parser.add_argument("--nodrop", action="store_true", help="whether to drop out the enhanced training dataset") parser.add_argument('--num_workers', type=int, default=8, help='number of workers in data loader') parser.add_argument('--regs', type=float, default=1e-5, help='Regularization.') parser.add_argument('--max2keep', type=int, default=1, help='max checkpoints to keep') args, _ = parser.parse_known_args() # InfoNCE if(args.model_name == 'InfoNCE'): parser.add_argument('--tau', type=float, default=0.1, help='temperature parameter') #AdvInfoNCE if(args.model_name == 'AdvInfoNCE'): parser.add_argument('--tau', type=float, default=0.1, help='temperature parameter') parser.add_argument('--eta_epochs', type=int, default=7, help='epochs for eta, control the disturbance of adv training') parser.add_argument('--adv_lr', type=float, default=5e-5, help='Learning rate for adversarial training.') parser.add_argument('--model_version', type=str, default='embed', help='model type, mlp or embed') parser.add_argument('--adv_interval',type=int,default=5, help='the interval of adversarial training') parser.add_argument('--warm_up_epochs', type=int, default=0, help='warm up epochs, in this stage, adv training is not used') parser.add_argument('--k_neg', type=float, default=64, help='k_neg for negative sampling') parser.add_argument('--adv_epochs',type=int,default=1, help='the epoch of adversarial training') parser.add_argument('--w_embed_size',type=int,default=64, help='dimension of weight embedding') # MultVAE if(args.model_name == 'MultVAE'): parser.add_argument('--total_anneal_steps', type=int, default=200000, help='total anneal steps') parser.add_argument('--anneal_cap', type=float, default=0.2, help='anneal cap') parser.add_argument('--p_dim0', type=int, default=200, help='p_dim0') parser.add_argument('--p_dim1', type=int, default=600, help='p_dim1') # AlphaRec if(('IntentCF' in args.model_name) or ('AlphaRec' in args.model_name)): parser.add_argument('--tau', type=float, default=0.1, help='temperature parameter') parser.add_argument('--lm_model', type=str, default='v3', choices=['bert', 'roberta', 'llama2_7b', 'llama3_7b', 'mistral_7b', 'v2', 'v3', 'SFR', 'v3_shuffle'], help='The base language model') parser.add_argument('--model_version', type=str, default='homo', choices=['mlp', 'homo'], help='The mapping method') if('LIntCF' in args.model_name): parser.add_argument('--tau', type=float, default=0.1, help='temperature parameter') parser.add_argument('--lambda_cl', type=float, default=1, help='Rate of contrastive loss') parser.add_argument('--temp_cl', type=float, default=0.15, help='Temperature of contrastive loss') parser.add_argument('--eps', type=float, default=0.1, help='Noise rate') #BC_LOSS if(args.model_name == 'BC_Loss'): parser.add_argument('--tau1', type=float, default=0.07, help='temperature parameter for L1') parser.add_argument('--tau2', type=float, default=0.1, help='temperature parameter for L2') parser.add_argument('--w_lambda', type=float, default=0.5, help='weight for combining l1 and l2.') parser.add_argument('--freeze_epoch',type=int,default=5) #SimpleX if(args.model_name == 'SimpleX'): parser.add_argument('--w_neg', type=float, default=1) parser.add_argument('--neg_margin',type=float, default=0.4) # SGL if(args.model_name == 'SGL'): parser.add_argument('--lambda_cl', type=float, default=0.2, help='Rate of contrastive loss') parser.add_argument('--temp_cl', type=float, default=0.15, help='Temperature of contrastive loss') parser.add_argument('--droprate', type=float, default=0.1, help='drop out rate for SGL') # XSimGCL & RLMRec & KAR if(args.model_name == "XSimGCL" or args.model_name == "RLMRec" or args.model_name == "KAR"): parser.add_argument('--lambda_cl', type=float, default=0.1, help='Rate of contrastive loss') parser.add_argument('--temp_cl', type=float, default=0.15, help='Temperature of contrastive loss') parser.add_argument('--layer_cl', type=int, default=1, help='Number of layers for contrastive loss') parser.add_argument('--eps_XSimGCL', type=float, default=0.1, help='Noise rate for XSimGCL') # RLMRec if(args.model_name == 'RLMRec'): parser.add_argument('--kd_temperature', type=float, default=0.2, help='temperature parameter') parser.add_argument('--kd_weight', type=float, default=1e-2, help='kd_weight') if(args.rs_type == 'Seq'): parser.add_argument('--r_click', type=float, default=0.2, help='reward for the click behavior.') parser.add_argument('--r_buy', type=float, default=1.0, help='reward for the purchase behavior.') parser.add_argument('--save_flag', type=int, default=0, help='0: Disable model saver, 1: Activate model saver') parser.add_argument('--loss', type=str, default='bpr', choices=['bpr', 'bce', 'mse'], help='loss function.') args, _ = parser.parse_known_args() # Model-specific Args if(args.model_name == 'Caser'): parser.add_argument('--num_filters', type=int, default=16, help='num_filters') parser.add_argument('--filter_sizes', nargs='?', default='[2,3,4]', help='Specify the filter_size') if(args.model_name == 'GRU4Rec'): parser.add_argument('--gru_layers', type=int, default=1, help='number of gru layers.') if(args.model_name == 'SASRec'): parser.add_argument('--num_heads', type=int, default=1, help='num_heads') if(args.rs_type == 'LLM'): parser.add_argument('--llm_path', type=str, default='meta-llama/Llama-2-7b-hf', help='path to llm model') parser.add_argument('--micro_batch_size', type=int, default=32, help='micro batch size') args, _ = parser.parse_known_args() if(args.model_name == 'RecInt'): parser.add_argument('--recommender', type=str, default='SASRec', help='SASRec, Caser, GRU4Rec, DreamRec') parser.add_argument('--rec_size', type=int, default=64, help='embedding size of the recommender') parser.add_argument('--rec_type', type=str, default='h_all', help='recommender type') parser.add_argument('--max_txt_len', type=int, default=32, help='max text length') parser.add_argument('--end_sym', type=str, default='\n', help='end symbol') # SASRec parser.add_argument('--num_heads', type=int, default=1, help='num_heads') if(args.model_name == 'TALLRec'): parser.add_argument('--sample_num', type=int, default=64, help='Sample number for training') parser.add_argument('--cutoff_len', type=int, default=512, help='cutoff length') parser.add_argument('--not_train_on_inputs', action="store_true", help='If True, masks out inputs in loss') parser.add_argument('--not_group_by_length', action="store_true", help='If False, groups samples by length') args_full, _ = parser.parse_known_args() special_args = list(set(vars(args_full).keys()) - set(vars(args).keys())) special_args.sort() if(args.model_name == 'UniSRec'): args_full.n_layers = 2 args_full.n_heads = 2 args_full.hidden_size = 300 args_full.plm_hidden_size = 768 args_full.inner_size = 256 args_full.hidden_dropout_prob = 0.5 args_full.attn_dropout_prob = 0.5 args_full.hidden_act = 'gelu' args_full.layer_norm_eps = 1e-12 args_full.initializer_range = 0.02 args_full.loss_type = 'CE' args_full.item_drop_ratio = 0.2 args_full.lambda_ = 1e-3 args_full.train_stage = 'pretrain' args_full.adaptor_dropout_prob = 0.2 args_full.adaptor_layers = [768, 300] args_full.temperature = 0.07 args_full.n_exps = 8 args_full.max_seq_length = 20 return args_full, special_args