-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathmodels.py
153 lines (121 loc) · 6.21 KB
/
models.py
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
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
import torch
import torch.nn.functional as F
import DiffModel as Diff
import sscdr_model as SSCDR
import lacdr_model as LACDR
class LookupEmbedding(torch.nn.Module):
def __init__(self, uid_all, iid_all, emb_dim):
super().__init__()
self.uid_embedding = torch.nn.Embedding(uid_all, emb_dim)
self.iid_embedding = torch.nn.Embedding(iid_all + 1, emb_dim)
def forward(self, x):
uid_emb = self.uid_embedding(x[:, 0].unsqueeze(1))
iid_emb = self.iid_embedding(x[:, 1].unsqueeze(1))
emb = torch.cat([uid_emb, iid_emb], dim=1)
return emb
class MetaNet(torch.nn.Module):
def __init__(self, emb_dim, meta_dim):
super().__init__()
self.event_K = torch.nn.Sequential(torch.nn.Linear(emb_dim, emb_dim), torch.nn.ReLU(),
torch.nn.Linear(emb_dim, 1, False))
self.event_softmax = torch.nn.Softmax(dim=1)
self.decoder = torch.nn.Sequential(torch.nn.Linear(emb_dim, meta_dim), torch.nn.ReLU(),
torch.nn.Linear(meta_dim, emb_dim * emb_dim))
def forward(self, emb_fea, seq_index):
mask = (seq_index == 0).float()
event_K = self.event_K(emb_fea)
t = event_K - torch.unsqueeze(mask, 2) * 1e8
att = self.event_softmax(t)
his_fea = torch.sum(att * emb_fea, 1)
output = self.decoder(his_fea)
return output.squeeze(1)
class MFBasedModel(torch.nn.Module):
def __init__(self, uid_all, iid_all, emb_dim, meta_dim_0):
super().__init__()
self.emb_dim = emb_dim
self.src_model = LookupEmbedding(uid_all, iid_all, emb_dim)
self.tgt_model = LookupEmbedding(uid_all, iid_all, emb_dim)
self.aug_model = LookupEmbedding(uid_all, iid_all, emb_dim)
self.meta_net = MetaNet(emb_dim, meta_dim_0)
self.mapping = torch.nn.Linear(emb_dim, emb_dim, False)
def forward(self, x, stage , device, diff_model=None, ss_model=None,la_model=None,is_task=False):
if stage == 'train_src':
emb = self.src_model.forward(x)
x = torch.sum(emb[:, 0, :] * emb[:, 1, :], dim=1)
return x
elif stage in ['train_tgt', 'test_tgt']:
emb = self.tgt_model.forward(x)
x = torch.sum(emb[:, 0, :] * emb[:, 1, :], dim=1)
return x
elif stage in ['train_aug', 'test_aug']:
emb = self.aug_model.forward(x)
x = torch.sum(emb[:, 0, :] * emb[:, 1, :], dim=1)
return x
elif stage in ['train_meta', 'test_meta']:
iid_emb = self.tgt_model.iid_embedding(x[:, 1].unsqueeze(1))
uid_emb_src = self.src_model.uid_embedding(x[:, 0].unsqueeze(1))
ufea = self.src_model.iid_embedding(x[:, 2:])
mapping = self.meta_net.forward(ufea, x[:, 2:]).view(-1, self.emb_dim, self.emb_dim)
uid_emb = torch.bmm(uid_emb_src, mapping)
emb = torch.cat([uid_emb, iid_emb], 1)
output = torch.sum(emb[:, 0, :] * emb[:, 1, :], dim=1)
return output
elif stage == 'train_map':
src_emb = self.src_model.uid_embedding(x.unsqueeze(1)).squeeze()
src_emb = self.mapping.forward(src_emb)
tgt_emb = self.tgt_model.uid_embedding(x.unsqueeze(1)).squeeze()
return src_emb, tgt_emb
elif stage == 'test_map':
uid_emb = self.mapping.forward(self.src_model.uid_embedding(x[:, 0].unsqueeze(1)).squeeze())
emb = self.tgt_model.forward(x)
emb[:, 0, :] = uid_emb
x = torch.sum(emb[:, 0, :] * emb[:, 1, :], dim=1)
return x
elif stage == 'train_diff':
tgt_uid, iid_input,y_input = x
tgt_emb = self.tgt_model.uid_embedding(tgt_uid.unsqueeze(1)).squeeze()
cond_emb = self.src_model.uid_embedding(tgt_uid.unsqueeze(1)).squeeze()
iid_emb = self.tgt_model.iid_embedding(iid_input.unsqueeze(1)).squeeze()
loss = Diff.diffusion_loss_fn(diff_model ,tgt_emb,cond_emb,\
iid_emb,y_input, device,is_task)
return loss
elif stage == 'test_diff':
tgt_uid, iid_input, _ = x
cond_emb = self.src_model.uid_embedding( tgt_uid.unsqueeze(1)).squeeze()
iid_emb = self.tgt_model.iid_embedding(iid_input.unsqueeze(1)).squeeze()
trans_emb,iid_emb_out = Diff.p_sample_loop( diff_model,cond_emb,iid_emb,device)
x = torch.sum( trans_emb * iid_emb_out, dim=1)
return x
elif stage == 'train_ss':
x_u, x_p_i,x_n_i,x_t_u = x
x_u_emb = self.src_model.uid_embedding(x_u.unsqueeze(1)).squeeze()
x_p_i_emb = self.src_model.iid_embedding( x_p_i.unsqueeze(1)).squeeze()
x_n_i_emb = self.src_model.iid_embedding( x_n_i.unsqueeze(1)).squeeze()
x_t_u_emb = self.tgt_model.uid_embedding(x_t_u.unsqueeze(1)).squeeze()
loss = SSCDR.sscdr_loss_fn(
ss_model.forward(x_u_emb),
ss_model.forward(x_p_i_emb),
ss_model.forward(x_n_i_emb),
x_t_u_emb
)
return loss
elif stage == 'test_ss':
uid_emb = ss_model.forward(self.src_model.uid_embedding(x[:, 0].unsqueeze(1)).squeeze())
emb = self.tgt_model.forward(x)
emb[:, 0, :] = uid_emb
x = torch.sum(emb[:, 0, :] * emb[:, 1, :], dim=1)
return x
elif stage == 'train_la':
x_uid, x_mask_src, x_mask_tgt = x
x_u_emb_s = self.src_model.uid_embedding(x_uid.unsqueeze(1)).squeeze()
x_u_emb_t = self.tgt_model.uid_embedding(x_uid.unsqueeze(1)).squeeze()
x_mask_src = x_mask_src.unsqueeze(1)
x_mask_tgt = x_mask_tgt.unsqueeze(1)
loss = LACDR.lacdr_loss_fn(la_model,x_u_emb_s,x_mask_src, x_u_emb_t,x_mask_tgt )
return loss
elif stage == 'test_la':
uid_emb = la_model.forward(self.src_model.uid_embedding(x[:, 0].unsqueeze(1)).squeeze())
emb = self.tgt_model.forward(x)
emb[:, 0, :] = uid_emb
x = torch.sum(emb[:, 0, :] * emb[:, 1, :], dim=1)
return x