-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcora-3.py
executable file
·127 lines (112 loc) · 6.81 KB
/
cora-3.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
import torch, pickle, sys, subprocess, random, copy, time, evaluate
import numpy as np
import importlib.util
from torch_geometric.data import Data, Batch
from data.handlers import subgraph_loader
from torch import nn
from layers.gnn import GraphSAGE
from external_libs.subgraph_utils import get_subgraph
from torch_geometric.nn import global_add_pool, global_mean_pool
from torch.nn import Sequential
import torch.multiprocessing
torch.multiprocessing.set_sharing_strategy('file_system')
""" srun -n 1 --gres=gpu:1 --cpus-per-task=15 --partition=ml-all-gpu --output=cora-3.txt python -u cora-3.py & """
""" srun -n 1 --gres=gpu:1 --cpus-per-task=15 --nice=1000 --partition=ml01-gpu,ml02-gpu,ml04-gpu,ml05-gpu,ml06-gpu,ml07-gpu,ml08-gpu --time=12:00:00 --pty bash """
""" Arguments: """
dataset = "cora"
k = 3
hidden_size = 128
tol = 1e-2
patience = 5
num_cpus = 10
lr = 0.001
num_tours = 80
batch_size = 50
super_node_size = 5000
cuda = True
train = { "DAG": torch.load("data/"+dataset+"/train.pt")[k]["DAG"] , "Hyperedge": torch.load("data/"+dataset+"/train.pt")[k]["Hyperedge"] }
test = { "DAG": torch.load("data/"+dataset+"/test.pt")[k]["DAG"] , "Hyperedge": torch.load("data/"+dataset+"/test.pt")[k]["Hyperedge"] }
graph = torch.load("data/"+dataset+"/data.pt")["graph"]
dataset = torch.load("data/"+dataset+"/data.pt")["dataset"]
input_size = graph.x.size(1)
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.gnn_layer = GraphSAGE( input_size, hidden_size, double_layer=False )
self.output_layer = nn.Sequential(nn.Linear(hidden_size, hidden_size), nn.LeakyReLU(), nn.Linear(hidden_size, 1))
def forward(self, data):
subgraph_data = subgraph_loader( data, k, super_node_size, num_tours, num_cpus )
subgraphs = [get_subgraph(data[subgraph_data.batch[i].item()], subgraph_data.subgraphs[i].squeeze()) for i in range(len(subgraph_data.subgraphs))]
subgraphs_lst = []
for i in range(0, len(subgraphs), 500):
subgraphs_b = Batch().from_data_list(subgraphs[i:i+min([500,len(subgraphs)-i])])
subgraphs_b = self.gnn_layer(subgraphs_b.x.cuda(), subgraphs_b.edge_index.cuda(), subgraphs_b.batch.cuda()) \
if next(self.parameters()).get_device() != -1 else self.gnn_layer(subgraphs_b.x, subgraphs_b.edge_index, subgraphs_b.batch)
subgraphs_lst.append(subgraphs_b)
subgraphs = torch.cat(subgraphs_lst,dim=0)
subgraphs = self.output_layer(subgraphs)
weights = subgraph_data.weights.cuda() if next(self.parameters()).get_device() != -1 else subgraph_data.weights
batch = subgraph_data.batch.cuda() if next(self.parameters()).get_device() != -1 else subgraph_data.batch
subgraphs = subgraphs*weights
norm = global_add_pool(weights, batch)
energy = global_add_pool(subgraphs, batch)
return energy/norm
def embedding(self, subgraphs):
with torch.no_grad():
subgraphs_lst = []
for i in range(0, len(subgraphs), 500):
subgraphs_b = Batch().from_data_list(subgraphs[i:i+min([500,len(subgraphs)-i])])
subgraphs_b = self.gnn_layer(subgraphs_b.x.cuda(), subgraphs_b.edge_index.cuda(), subgraphs_b.batch.cuda()) \
if next(self.parameters()).get_device() != -1 else self.gnn_layer(subgraphs_b.x, subgraphs_b.edge_index, subgraphs_b.batch)
subgraphs_lst.append(subgraphs_b)
subgraphs = torch.cat(subgraphs_lst,dim=0)
return subgraphs
def init_weights(m):
if type(m) == torch.nn.Linear:
torch.nn.init.xavier_uniform_(m.weight)
""" Creating samplers to parallelize over each minibatch """
for i in range(1,batch_size+1):
process = subprocess.Popen("cp external_libs/rgpm/tradicional external_libs/rgpm/tradicional-"+str(i), stdout=subprocess.PIPE, stderr=None, shell=True, stdin=subprocess.PIPE)
process.wait()
model = Net()
model.apply(init_weights)
best_model = copy.deepcopy(model)
best_loss = float("inf")
cn_wait = 0
step = 0
print("===\t RANDOM MODEL RESULTS \t===")
print("dag:\t", evaluate.logit(train_x=best_model.embedding([ get_subgraph(graph, train["DAG"]["x"][i].squeeze()) for i in range(len(train["DAG"]["x"])) ]).detach().cpu(), train_y=train["DAG"]["y"], \
test_x=best_model.embedding([ get_subgraph(graph, test["DAG"]["x"][i].squeeze()) for i in range(len(test["DAG"]["x"])) ]).detach().cpu(), test_y=test["DAG"]["y"] ), \
"hyperedge:\t", evaluate.logit(train_x=best_model.embedding([ get_subgraph(graph, train["Hyperedge"]["x"][i].squeeze()) for i in range(len(train["Hyperedge"]["x"])) ]).detach().cpu(), train_y=train["Hyperedge"]["y"], \
test_x=best_model.embedding([ get_subgraph(graph, test["Hyperedge"]["x"][i].squeeze()) for i in range(len(test["Hyperedge"]["x"])) ]).detach().cpu(), test_y=test["Hyperedge"]["y"] ) )
print("===\t \t===")
if cuda: model = model.cuda()
optimizer = torch.optim.Adam(model.parameters(), lr=lr)
loss_fn = torch.nn.BCEWithLogitsLoss()
random.shuffle(dataset)
stime = time.time()
for i in range(0, len(dataset), batch_size):
model.train()
optimizer.zero_grad()
y = torch.tensor([graph.y for graph in dataset[i:i+min([batch_size,len(dataset)-i])]]).float().unsqueeze(-1)
if cuda: y = y.cuda()
energy = model(dataset[i:i+min([batch_size,len(dataset)-i])])
loss = loss_fn(-energy, y)
loss.backward()
optimizer.step()
if loss.item() <= best_loss - tol:
best_loss, cn_wait = loss.item(), 1
best_model = copy.deepcopy(model)
else:
cn_wait += 1
print("loss:\t", loss.item(), "time:\t", time.time()-stime, "step:\t", step, "dag:\t", evaluate.logit(train_x=model.embedding([ get_subgraph(graph, train["DAG"]["x"][i].squeeze()) \
for i in range(len(train["DAG"]["x"])) ]).detach().cpu(), train_y=train["DAG"]["y"], \
test_x=model.embedding([ get_subgraph(graph, test["DAG"]["x"][i].squeeze()) for i in range(len(test["DAG"]["x"])) ]).detach().cpu(), test_y=test["DAG"]["y"] ))
if cn_wait == patience and best_loss < 0.2: break
step += 1
print("===\t TRAINED MODEL RESULTS \t===")
print("dag:\t", evaluate.logit(train_x=best_model.embedding([ get_subgraph(graph, train["DAG"]["x"][i].squeeze()) for i in range(len(train["DAG"]["x"])) ]).detach().cpu(), train_y=train["DAG"]["y"], \
test_x=best_model.embedding([ get_subgraph(graph, test["DAG"]["x"][i].squeeze()) for i in range(len(test["DAG"]["x"])) ]).detach().cpu(), test_y=test["DAG"]["y"] ), \
"hyperedge:\t", evaluate.logit(train_x=best_model.embedding([ get_subgraph(graph, train["Hyperedge"]["x"][i].squeeze()) for i in range(len(train["Hyperedge"]["x"])) ]).detach().cpu(), train_y=train["Hyperedge"]["y"], \
test_x=best_model.embedding([ get_subgraph(graph, test["Hyperedge"]["x"][i].squeeze()) for i in range(len(test["Hyperedge"]["x"])) ]).detach().cpu(), test_y=test["Hyperedge"]["y"] ) )
print("===\t \t===")