forked from Nicoleqwerty/RDF-to-Text
-
Notifications
You must be signed in to change notification settings - Fork 0
/
main_PG.py
228 lines (179 loc) · 10.1 KB
/
main_PG.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
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
import os
import time
# import yaml
import json
import random
import torch
import torch.optim as optim
from collections import namedtuple
import math
import sys
from data_loader import get_loader, get_gcn_loader, get_gcn_lstm_loader, \
get_gtr_loader, get_gcn_gtr_loader
from torch_geometric.data import DataLoader
from LSTMModel import LSTMModel
from GCNModel import GCNModel
from GTRModel import GTRModel2
from utils import config
from GCNGTR2Model import GCNGTR2Model
import numpy as np
device_id=0
SEED = 43
# SEED = 1234
random.seed(SEED)
torch.manual_seed(SEED)
torch.backends.cudnn.deterministic = True
# torch.cuda.set_device(device_id)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# device = 'cpu'
model_type = config.model_type
def count_parameters(model):
return sum(p.numel() for p in model.parameters() if p.requires_grad)
def epoch_time(start_time, end_time):
elapsed_time = end_time - start_time
elapsed_mins = int(elapsed_time / 60)
elapsed_secs = int(elapsed_time - elapsed_time * 60)
return elapsed_mins, elapsed_secs
def train(model, iterator, optimizer, epoch, id2word):
if model_type == 'gcn_lstm':
model.encoder_gcn.train()
model.encoder_lstm.train()
elif model_type == 'gcn_gtr2':
model.encoder_gcn.train()
model.encoder_gtr.train()
elif model_type == 'gcn_gtr2_lstm':
model.encoder_gcn.train()
model.encoder_gtr.train()
model.encoder_lstm.train()
elif model_type == 'gcn_gtr2_ae':
model.encoder_gcn.train()
model.encoder_gtr.train()
model.encoder_ae.train()
else:
model.encoder.train()
model.decoder.train()
model.reduce_state.train()
return model.forward(iterator, device, optimizer, epoch)
def evaluate(model, iterator):
if model_type == 'gcn_gtr2':
model.encoder_gcn.eval()
model.encoder_gtr.eval()
else:
model.encoder.eval()
model.decoder.eval()
model.reduce_state.eval()
with torch.no_grad():
epoch_loss = model.eval(iterator, device)
return epoch_loss/len(iterator)
def sort_beams( beams):
return sorted(beams, key=lambda h: h.avg_log_prob, reverse=True)
def beam_search(model, iterator, word2id, id2word, rplc_dict=None):
if model_type=='gcn_gtr2':
model.encoder_gcn.eval()
model.encoder_gtr.eval()
else:
model.encoder.eval()
model.decoder.eval()
model.reduce_state.eval()
translate_results = []
f = open(config.test_write_to, 'w+')
if rplc_dict==None:
with torch.no_grad():
model.beam_search(f, iterator,device,word2id,id2word)
else:
with torch.no_grad():
model.beam_search(f, iterator,device,word2id,id2word,rplc_dict)
f.close()
def main():
with open(config.word2id_vocab_file, 'r', encoding = 'utf-8') as fd:
word2id = json.load(fd)
with open(config.id2word_vocab_file, 'r', encoding = 'utf-8') as fd:
id2word = json.load(fd)
if model_type == 'bilstm':
model = LSTMModel()
params = list(model.encoder.parameters()) + list(model.decoder.parameters()) + \
list(model.reduce_state.parameters())
data_loader_train = get_loader(config.train_data_file, word2id, max_enc_steps=config.max_enc_steps, max_dec_steps=config.max_dec_steps, batch_size=config.batch_size)
data_loader_val = get_loader(config.val_data_file, word2id, max_enc_steps=config.max_enc_steps, max_dec_steps=config.max_dec_steps,batch_size=config.batch_size)
data_loader_test = get_loader(config.test_data_file, word2id, max_enc_steps=config.max_enc_steps, max_dec_steps=config.max_dec_steps,batch_size=1, shuffle=False)
elif model_type == 'gcn':
model = GCNModel()
params = list(model.encoder.parameters()) + list(model.decoder.parameters()) + list(model.reduce_state.parameters())
data_loader_train = get_gcn_loader(config.train_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=config.batch_size)
data_loader_val = get_gcn_loader(config.val_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=config.batch_size)
data_loader_test = get_gcn_loader(config.test_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=1, shuffle=False)
elif model_type == 'gtr' or model_type == 'gtr2':
model = GTRModel2()
# model = GTRModel()
params = list(model.encoder.parameters()) + list(model.decoder.parameters()) + list(model.reduce_state.parameters())
data_loader_train = get_gtr_loader(config.train_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=config.batch_size)
data_loader_val = get_gtr_loader(config.val_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=config.batch_size)
data_loader_test = get_gtr_loader(config.test_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=1, shuffle=False)
elif model_type == 'gcn_gtr2':
model = GCNGTR2Model()
params = list(model.encoder_gcn.parameters()) + list(model.encoder_gtr.parameters()) + \
list(model.decoder.parameters()) + list(model.reduce_state.parameters())
data_loader_train = get_gcn_gtr_loader(config.train_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=config.batch_size)
data_loader_val = get_gcn_gtr_loader(config.val_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=config.batch_size)
data_loader_test = get_gcn_gtr_loader(config.test_data_file, word2id, max_enc_steps=config.max_enc_steps,
max_dec_steps=config.max_dec_steps, batch_size=1, shuffle=False)
num_nodes_batch = []
for i, (src_inputs, trg_seqs, trg_lengths) in enumerate(data_loader_train):
src_gcn_inputs = src_inputs['src_gcn_inputs']
src_gcn_inputs = list(DataLoader(src_gcn_inputs, len(src_gcn_inputs)))[0].to(device)
num_nodes_batch.append(src_gcn_inputs.num_nodes)
initial_lr = config.lr_coverage if config.is_coverage else config.lr
optimizer = optim.Adam(params, lr=initial_lr)
# optimizer = optim.Adam(params, lr=initial_lr, initial_accumulator_value=config.adagrad_init_acc)
# best_valid_loss = 1.19
best_valid_loss = float('inf')
# if load_model:
# if model_type=='gcn_lstm':
# model.encoder_gcn.load_state_dict(torch.load(config.save_model_path + '_encoder_gcn', map_location={'cuda:1':'cuda:3'}))
# model.encoder_lstm.load_state_dict(torch.load(config.save_model_path + '_encoder_lstm', map_location={'cuda:1':'cuda:3'}))
# else:
# model.encoder.load_state_dict(torch.load(config.save_model_path + '_encoder'))
#
# model.decoder.load_state_dict(torch.load(config.save_model_path + '_decoder', map_location={'cuda:1':'cuda:3'}))
# model.reduce_state.load_state_dict(torch.load(config.save_model_path + '_reduce_state', map_location={'cuda:1':'cuda:3'}))
if config.load_model:
if model_type=='gcn_gtr2':
# model.encoder_gcn.load_state_dict(torch.load(config.save_model_path + '_encoder_gcn'))
model.encoder_gcn.load_state_dict(torch.load(config.save_model_path + '_encoder_gcn', map_location=torch.device('cpu')))
# model.encoder_gtr.load_state_dict(torch.load(config.save_model_path + '_encoder_gtr'))
model.encoder_gtr.load_state_dict(torch.load(config.save_model_path + '_encoder_gtr', map_location=torch.device('cpu')))
else:
model.encoder.load_state_dict(torch.load(config.save_model_path + '_encoder', map_location='cpu'))
# model.decoder.load_state_dict(torch.load(config.save_model_path + '_decoder'))
model.decoder.load_state_dict(torch.load(config.save_model_path + '_decoder', map_location=torch.device('cpu')))
# model.reduce_state.load_state_dict(torch.load(config.save_model_path + '_reduce_state'))
model.reduce_state.load_state_dict(torch.load(config.save_model_path + '_reduce_state', map_location=torch.device('cpu')))
if config.mode == 'train':
for epoch in range(config.n_epoch):
train_loss = train(model, data_loader_train, optimizer, epoch=epoch, id2word=id2word)
valid_loss = evaluate(model, data_loader_val)
print('Train Loss: {} | Train PPL: {}'.format(train_loss, math.exp(train_loss)))
print('Val. Loss: {} | Val. PPL: {}'.format(valid_loss, math.exp(valid_loss)))
if valid_loss < best_valid_loss:
best_valid_loss = valid_loss
if model_type=='gcn_gtr2':
torch.save(model.encoder_gcn.state_dict(), config.save_model_path+'_encoder_gcn')
torch.save(model.encoder_gtr.state_dict(), config.save_model_path+'_encoder_gtr')
else:
torch.save(model.encoder.state_dict(), config.save_model_path + '_encoder')
torch.save(model.decoder.state_dict(), config.save_model_path+'_decoder')
torch.save(model.reduce_state.state_dict(), config.save_model_path+'_reduce_state')
elif config.mode == 'translate':
# beam_search(model, data_loader_test, word2id, id2word, rplc_dict)
beam_search(model, data_loader_test, word2id, id2word)
if __name__ == '__main__':
main()
# nohup python main_PG.py > log/GCN_l2_seed43.log 2>&1 &