-
Notifications
You must be signed in to change notification settings - Fork 7
/
Copy pathtraingnn_reproduce.sh
135 lines (94 loc) · 17 KB
/
traingnn_reproduce.sh
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
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
DATASET=huggingface
# GCN
# - LM Freeze # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GCN', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GCN', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --lm_frozen=1 --epoch=10 --text_negative=1 --gnn_name=GCN --lr=0.001
python main.py --lm_frozen=0 --epoch=20 --text_negative=1 --gnn_name=GCN
# GAT
# - LM Freeze # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GAT', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=0, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GAT', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --lm_frozen=1 --epoch=10 --text_negative=0 --gnn_name=GAT --lr=0.001
python main.py --lm_frozen=0 --epoch=20 --text_negative=1 --gnn_name=GAT
# SAGE
# - LM Freeze # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SAGE', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SAGE', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --lm_frozen=1 --epoch=10 --text_negative=1 --gnn_name=SAGE --lr=0.001
python main.py --lm_frozen=0 --epoch=20 --text_negative=1 --gnn_name=SAGE
# GIN
# - LM Freeze # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GIN', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=0, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GIN', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --lm_frozen=1 --epoch=10 --text_negative=0 --gnn_name=GIN --lr=0.001
python main.py --lm_frozen=0 --epoch=20 --text_negative=1 --gnn_name=GIN
# TransformerConv
# - LM Freeze # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='TransformerConv', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='TransformerConv', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --lm_frozen=1 --epoch=10 --text_negative=1 --gnn_name=TransformerConv --lr=0.001
python main.py --lm_frozen=0 --epoch=20 --text_negative=1 --gnn_name=TransformerConv
# (重新训练的时候co-train用的epoch=10)
# SGC
# - LM Freeze # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SGC', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='huggingface', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SGC', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --lm_frozen=1 --epoch=10 --text_negative=1 --gnn_name=SGC --lr=0.001
python main.py --lm_frozen=0 --epoch=20 --text_negative=1 --gnn_name=SGC
DATASET=dailylife
# GCN
# - LM Freeze # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GCN', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=6, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GCN', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=0, train_num=3000)
python main.py --dataset=$DATASET --lm_frozen=1 --epoch=10 --gnn_name=GCN --text_negative=1 --lr=0.001
python main.py --dataset=$DATASET --lm_frozen=0 --epoch=6 --gnn_name=GCN --text_negative=0
# GAT
# - LM Freeze # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GAT', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=6, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GAT', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=0, train_num=3000)
python main.py --dataset=$DATASET --lm_frozen=1 --epoch=10 --gnn_name=GAT --text_negative=1 --lr=0.001
python main.py --dataset=$DATASET --lm_frozen=0 --epoch=6 --gnn_name=GAT --text_negative=0
# SAGE
# - LM Freeze # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SAGE', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=6, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SAGE', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --dataset=$DATASET --lm_frozen=1 --epoch=10 --gnn_name=SAGE --text_negative=1 --lr=0.001
python main.py --dataset=$DATASET --lm_frozen=0 --epoch=6 --gnn_name=SAGE --text_negative=1
# GIN
# - LM Freeze # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GIN', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=0, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=6, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GIN', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=0, train_num=3000)
python main.py --dataset=$DATASET --lm_frozen=1 --epoch=10 --gnn_name=GIN --text_negative=0 --lr=0.001
python main.py --dataset=$DATASET --lm_frozen=0 --epoch=6 --gnn_name=GIN --text_negative=0
# TransformerConv
# - LM Freeze # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='TransformerConv', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=128, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=6, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='TransformerConv', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=0, train_num=3000)
python main.py --dataset=$DATASET --lm_frozen=1 --epoch=10 --gnn_name=TransformerConv --text_negative=1 --lr=0.001
python main.py --dataset=$DATASET --lm_frozen=0 --epoch=6 --gnn_name=TransformerConv --text_negative=0 --batch_size=128
# SGC
# - LM Freeze # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SGC', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=0, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='dailylife', device='cuda:0', dynamic_sample=0, epoch=6, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SGC', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --dataset=$DATASET --lm_frozen=1 --epoch=10 --gnn_name=SGC --text_negative=0 --lr=0.001
python main.py --dataset=$DATASET --lm_frozen=0 --epoch=6 --gnn_name=SGC --text_negative=1
DATASET=multimedia
# GCN
# - LM Freeze # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GCN', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GCN', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --dataset=$DATASET --epoch=10 --lm_frozen=1 --gnn_name=GCN --text_negative=1 --lr=1e-3
python main.py --dataset=$DATASET --epoch=20 --lm_frozen=0 --gnn_name=GCN --text_negative=1
# GAT
# - LM Freeze # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GAT', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GAT', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --dataset=$DATASET --epoch=10 --lm_frozen=1 --gnn_name=GAT --text_negative=1 --lr=1e-3
python main.py --dataset=$DATASET --epoch=20 --lm_frozen=0 --gnn_name=GAT --text_negative=1
# SAGE
# - LM Freeze # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SAGE', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=0, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SAGE', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=0, train_num=3000)
python main.py --dataset=$DATASET --epoch=10 --lm_frozen=1 --gnn_name=SAGE --text_negative=0 --lr=1e-3
python main.py --dataset=$DATASET --epoch=20 --lm_frozen=0 --gnn_name=SAGE --text_negative=0
# GIN
# - LM Freeze # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GIN', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=0, train_num=3000)
# - lM Tune # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='GIN', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --dataset=$DATASET --epoch=10 --lm_frozen=1 --gnn_name=GIN --text_negative=0 --lr=1e-3
python main.py --dataset=$DATASET --epoch=20 --lm_frozen=0 --gnn_name=GIN --text_negative=1
# TransformerConv
# - LM Freeze # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='TransformerConv', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=0, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='TransformerConv', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --dataset=$DATASET --epoch=10 --lm_frozen=1 --gnn_name=TransformerConv --text_negative=0 --lr=1e-3
python main.py --dataset=$DATASET --epoch=20 --lm_frozen=0 --gnn_name=TransformerConv --text_negative=1
# SGC
# - LM Freeze # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=10, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SGC', lm_frozen=1, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=0.001, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=1, seed=0, text_negative=1, train_num=3000)
# - LM Tune # Namespace(batch_size=512, dataset='multimedia', device='cuda:0', dynamic_sample=0, epoch=20, gnn_hidden_dim=1024, gnn_layer=1, gnn_name='SGC', lm_frozen=0, lm_name='intfloat/e5-large', load_alignment=True, load_model=0, lr=2e-05, maximum='', measure='dot', no_gnn_ablation=1, num_negatives=2, patience=5, save_model=0, seed=0, text_negative=1, train_num=3000)
python main.py --dataset=$DATASET --epoch=10 --lm_frozen=1 --gnn_name=SGC --text_negative=1 --lr=1e-3
python main.py --dataset=$DATASET --epoch=20 --lm_frozen=0 --gnn_name=SGC --text_negative=1