-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathTrainer.py
117 lines (83 loc) · 4.29 KB
/
Trainer.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
import torch
import torch.nn as nn
import torch.nn.functional as F
from utils import eval_Z
from torch_geometric.utils import structured_negative_sampling
class Trainer(object):
def __init__(self, model, X, edge_index, args):
self.model = model
self.lr = args.lr
self.optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, self.model.parameters()),
lr=self.lr,
weight_decay=1e-8)
self.bestResult = -1
self.beta = args.beta
self.gamma = args.gamma
self.sigma = args.sigma
self.X = X
self.edge_index = edge_index
self.node_num, self.attr_num = X.shape
self.sim_loss_m = args.m
def train_mini_batch(self):
self.optimizer.zero_grad()
loss.backward()
self.optimizer.step()
return loss.item()
def tf_loss(self):
attr_emb = self.model.encoder(torch.eye(self.attr_num).to(self.X.device))
return (attr_emb.norm(dim=1) * self.X.sum(dim=0)).mean()
def ae_loss(self):
# return (self.X - self.model.decoder(self.model.encoder(self.X))).norm()/self.X.shape[0]
return F.mse_loss(self.X, self.model.decoder(self.model.encoder(self.X)))
def sim_loss(self):
node_emb = self.model.forward(self.X, self.edge_index)
src_nodes, pos_nodes, neg_nodes = structured_negative_sampling(self.edge_index)
edges = torch.cat([torch.stack([src_nodes, pos_nodes]),torch.stack([src_nodes, neg_nodes])],dim=-1)
# sim_label= torch.norm(self.dist_M[src_nodes]-self.dist_M[pos_nodes],dim=1)<torch.norm(self.dist_M[src_nodes]-self.dist_M[neg_nodes],dim=1)
sim_label= torch.norm(self.X[src_nodes]-self.X[pos_nodes],dim=1)<torch.norm(self.X[src_nodes]-self.X[neg_nodes],dim=1)
sim_loss = (node_emb[src_nodes]*node_emb[pos_nodes]).sum(dim=1) - (node_emb[src_nodes]*node_emb[neg_nodes]).sum(dim=1)
sim_label = sim_label*2-1
return F.relu(sim_loss * -sim_label + self.sim_loss_m).mean()
def sage_sim_loss(self):
node_emb = self.model.sage_forward(self.X, self.edge_index)
src_nodes, pos_nodes, neg_nodes = structured_negative_sampling(self.edge_index)
edges = torch.cat([torch.stack([src_nodes, pos_nodes]),torch.stack([src_nodes, neg_nodes])],dim=-1)
# sim_label= torch.norm(self.dist_M[src_nodes]-self.dist_M[pos_nodes],dim=1)<torch.norm(self.dist_M[src_nodes]-self.dist_M[neg_nodes],dim=1)
sim_label= torch.norm(self.X[src_nodes]-self.X[pos_nodes],dim=1)<torch.norm(self.X[src_nodes]-self.X[neg_nodes],dim=1)
sim_loss = (node_emb[src_nodes]*node_emb[pos_nodes]).sum(dim=1) - (node_emb[src_nodes]*node_emb[neg_nodes]).sum(dim=1)
sim_label = sim_label*2-1
return F.relu(sim_loss * -sim_label + self.sim_loss_m).mean()
def train_batch(self):
self.model.train()
self.optimizer.zero_grad()
loss = self.sigma*self.ae_loss() + self.beta*self.tf_loss() + self.gamma*self.sim_loss()
loss.backward()
self.optimizer.step()
return loss.item()
def train_sage(self):
self.model.train()
self.optimizer.zero_grad()
loss = self.gamma*self.sage_sim_loss()
loss.backward()
self.optimizer.step()
return loss.item()
def test(self, qX, ans, topk, verbose=False):
self.model.eval()
node_emb = self.model.forward(self.X, self.edge_index)
q_emb = self.model.encoder(qX)
avg_hit = eval_Z(node_emb, q_emb, ans, k=topk, verbose=verbose)
return avg_hit
def sage_test(self, qX, ans, topk, verbose=False):
self.model.eval()
node_emb = self.model.sage_forward(self.X, self.edge_index)
q_emb = self.model.encoder(qX)
avg_hit = eval_Z(node_emb, q_emb, ans, k=topk, verbose=verbose)
return avg_hit
def save(self, dir):
if dir is not None:
torch.save(self.model.state_dict(), dir)
def decay_learning_rate(self, epoch, init_lr):
lr = init_lr / (1 + 0.05 * epoch)
for param_group in self.optimizer.param_groups:
param_group['lr'] = lr
return self.optimizer