UnlearnRec/params.py at main · HKUDS/UnlearnRec · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
import argparse


def str2bool(v):
    if isinstance(v, bool):
        return v
    if v.lower() in ('yes', 'true', 't', 'y', '1'):
        return True
    elif v.lower() in ('no', 'false', 'f', 'n', '0'):
        return False
    else:
        raise argparse.ArgumentTypeError('Boolean value expected.')


def ParseArgs():
	parser = argparse.ArgumentParser(description='Model Params')

	parser.add_argument('--gpu', default='0', type=str, help='indicates which gpu to use')
	parser.add_argument('--seed', default=1234, type=int, help='random seed')


	# Training
	parser.add_argument('--lr', default=1e-3, type=float, help='learning rate')
	parser.add_argument('--batch', default=512, type=int, help='batch size; key parameter.')
	# parser.add_argument('--batch', default=4096, type=int, help='batch size')
    # parser.add_argument('--batch', default=8192, type=int, help='batch size')


	parser.add_argument('--tst_bat', default=256, type=int, help='number of users in a testing batch')
	parser.add_argument('--epoch', default=200, type=int, help='number of total training epochs')
	parser.add_argument('--sim_epoch', default=5, type=int, help='number of simulated epochs')
	parser.add_argument('--decay', default=1.0, type=float, help='weight decay rate')


	# Checkpointing
	parser.add_argument('--early_stop', default=10, type=int, help='number of epochs for early stop')
	parser.add_argument('--model_2_finetune', default='./checkpoints/gowalla/unlearn/NOadv_unlearn_gowalla_simgcl_pdrp5p_tdrp3p_bpr1_unlearn1_align_t10_1_leaky0.99_unssl1e-3', help='path of pretrained (for unlearning) model to finetune.')
	parser.add_argument('--trained_model', default="./checkpoints/gowalla/before_unlearning/pretrain_gowalla_simgcl_advlightgcn0.5_reg1e-7_lr1e-3_ssl1e-2_esp2e-1_t1e-1_b4096_ep200_dim128_ly3", help='name of trained model to unlearn.')
	parser.add_argument('--save_path', default='./checkpoints/gowalla/pretrain_4_unlearning/tmp', help='file name to save model.')


	# Testing
	parser.add_argument('--topk', default=20, type=int, help='K of top K')
	# parser.add_argument('--data', default='ml1m', type=str, help='name of dataset')
	# parser.add_argument('--data', default='yelp2018', type=str, help='name of dataset')
	parser.add_argument('--data', default='gowalla', type=str, help='name of dataset')
	parser.add_argument('--tst_epoch', default=3, type=int, help='number of epoch to test while training')


	# Neural structure
	parser.add_argument('--latdim', default=128, type=int, help='embedding size')
	parser.add_argument('--gnn_layer', default=3, type=int, help='number of gnn layers')
	parser.add_argument('--unlearn_layer', default=0, type=int, help='number of unlearning gnn layers for unlearning neural network. 0 by default.')


	# Model type
	parser.add_argument('--model', default='simgcl', type=str, help='the model type to unlearn')
	# parser.add_argument('--model', default='lightgcn', type=str, help='the model type to unlearn')
	# parser.add_argument('--model', default='sgl', type=str, help='the model type to unlearn')

	# Exclusive for Simgcl Models
	parser.add_argument('--ssl_reg', default=1e-2, type=float, help='weight decay regularizer')
	parser.add_argument('--eps', default=0.2, type=float, help='epsilon in the model')
	parser.add_argument('--temp', default=0.1, type=float, help='temperature in ssl')
	parser.add_argument('--graphSampleN', default=15000, type=int, help='number of nodes to sample each time')
	parser.add_argument('--noiseRate', default=-0.1, type=float, help='ratio of edges to add noise')
	parser.add_argument('--reg_version', default='v1', type=str, help='the version of reg loss')

	# Exclusive for SGL Models
	parser.add_argument('--sgl_ssl_reg', default=1e-2, type=float, help='weight decay regularizer')
	parser.add_argument('--sgltemp', default=0.1, type=float, help='temperature in ssl')
	parser.add_argument('--sglkeepRate', default=0.8, type=float, help='temperature in ssl')


	# Loss weights
	parser.add_argument('--reg', default=1e-7, type=float, help='weight decay regularizer')
	parser.add_argument('--bpr_wei', default=1., type=float, help='weight for BPR loss')
	parser.add_argument('--unlearn_wei', default=0.5, type=float, help='weight for unlearning loss')
	parser.add_argument('--align_wei', default=0.02, type=float, help='weight for alignment loss')
	parser.add_argument('--align_temp', default=10., type=float, help='temperature for infoNCE')
	parser.add_argument('--align_type', default='v2', type=str, help='version for aligment function')
	parser.add_argument('--unlearn_type', default='v1', type=str, help='version for unlearning function')
	parser.add_argument('--perf_degrade', default=0.5, type=float, help='acceptable level of performance degradation')


	# Datasets for adversarial attack
	parser.add_argument('--adversarial_attack', default=False, type=str2bool, help='whether to use the datasets with adversarial attack')
	parser.add_argument('--random_attack', default=False, type=str2bool, help='whether to use the datasets with random adversarial attack. False by default.')
	parser.add_argument('--adv_method', default='lightgcn0.5', type=str, help='version for adversarial datasets.')
	# parser.add_argument('--adv_method', default='lightgcn', type=str, help='version for adversarial datasets')


	# Unlearn
	parser.add_argument('--overall_withdraw_rate', default=0.1, type=float, help='overall withdraw rate')
	parser.add_argument('--withdraw_rate_init', default=1, type=float, help='overall withdraw vector initialization')
	parser.add_argument('--allgrad', default=True, type=str2bool, help='whether to update all the parameters when training.')
	parser.add_argument('--pretrain_drop_rate', default=0.1, type=float, help='ratio of edges to drop when pretraining for unlearning.')
	parser.add_argument('--test_drop_rate', default=0.003, type=float, help='ratio of edges to drop when testing, only take effect when adversarial_attack is False')
	parser.add_argument('--hyper_temp', default=1., type=float, help='temperature for unlearning ssl')
	parser.add_argument('--unlearn_ssl', default=1e-3, type=float, help='weight for unlearning ssl')
	# parser.add_argument('--keep_rate', default=1., type=float, help='ratio of edges to keep, deprecated')
	parser.add_argument('--act', default='leaky', type=str, help='activation function for unlearnning fine-tune part')
	parser.add_argument('--layer_mlp', default=2, type=int, help='layer num for mlp (e.g., unlearning fine-tuning part)')
	parser.add_argument('--leaky', default=0.99, type=float, help='slope for the negative part of leaky relu')

	# parser.add_argument('--keepRate', default=1.0, type=float, help='ratio of edges to keep')


	parser.add_argument('--fineTune', default=True, type=str2bool, help='whether to start finetuning.')


	return parser.parse_args()


args = ParseArgs()