forked from PaddlePaddle/PGL
-
Notifications
You must be signed in to change notification settings - Fork 0
/
node_classification_infer.py
188 lines (158 loc) · 6.08 KB
/
node_classification_infer.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
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import division
from __future__ import absolute_import
from __future__ import print_function
from __future__ import unicode_literals
import argparse
import pickle
import time
import glob
import os
import io
import traceback
import pickle as pkl
role = os.getenv("TRAINING_ROLE", "TRAINER")
import numpy as np
import yaml
from easydict import EasyDict as edict
import pgl
from pgl.utils.logger import log
from pgl.utils import paddle_helper
import paddle
import paddle.fluid as F
from models.model import NodeClassificationModel
from dataset.graph_reader import NodeClassificationGenerator
class PredictData(object):
def __init__(self, num_nodes):
trainer_id = int(os.getenv("PADDLE_TRAINER_ID", "0"))
trainer_count = int(os.getenv("PADDLE_TRAINERS_NUM", "1"))
train_usr = np.arange(trainer_id, num_nodes, trainer_count)
#self.data = (train_usr, train_usr)
self.data = train_usr
def __getitem__(self, index):
return [self.data[index], self.data[index]]
def tostr(data_array):
return " ".join(["%.5lf" % d for d in data_array])
def run_predict(py_reader,
exe,
program,
model_dict,
log_per_step=1,
args=None):
id2str = io.open(os.path.join(args.graph_work_path, "terms.txt"), encoding=args.encoding).readlines()
trainer_id = int(os.getenv("PADDLE_TRAINER_ID", "0"))
trainer_count = int(os.getenv("PADDLE_TRAINERS_NUM", "1"))
if not os.path.exists(args.output_path):
os.mkdir(args.output_path)
fout = io.open("%s/part-%s" % (args.output_path, trainer_id), "w", encoding="utf8")
batch = 0
for batch_feed_dict in py_reader():
batch += 1
_, batch_node_real_index, batch_logits = exe.run(
program,
feed=batch_feed_dict,
fetch_list=model_dict.outputs)
if batch % log_per_step == 0:
log.info("Predict %s finished" % batch)
for idx, logits in zip(batch_node_real_index, batch_logits):
if args.input_type == "text":
text = id2str[int(idx)].strip("\n").split("\t")[-1]
#prediction = np.argmax(logits)
prediction = logits[1]
line = "{}\t{}\n".format(text, prediction)
fout.write(line)
fout.close()
def _warmstart(exe, program, path='params'):
def _existed_persitables(var):
#if not isinstance(var, fluid.framework.Parameter):
# return False
if not F.io.is_persistable(var):
return False
param_path = os.path.join(path, var.name)
log.info("Loading parameter: {} persistable: {} exists: {}".format(
param_path,
F.io.is_persistable(var),
os.path.exists(param_path),
))
return os.path.exists(param_path)
F.io.load_vars(
exe,
path,
main_program=program,
predicate=_existed_persitables
)
def main(config):
model = NodeClassificationModel(config)
if config.learner_type == "cpu":
place = F.CPUPlace()
elif config.learner_type == "gpu":
gpu_id = int(os.getenv("FLAGS_selected_gpus", "0"))
place = F.CUDAPlace(gpu_id)
else:
raise ValueError
exe = F.Executor(place)
val_program = F.default_main_program().clone(for_test=True)
exe.run(F.default_startup_program())
_warmstart(exe, F.default_startup_program(), path=config.infer_model)
num_threads = int(os.getenv("CPU_NUM", 1))
trainer_id = int(os.getenv("PADDLE_TRAINER_ID", 0))
exec_strategy = F.ExecutionStrategy()
exec_strategy.num_threads = num_threads
build_strategy = F.BuildStrategy()
build_strategy.enable_inplace = True
build_strategy.memory_optimize = True
build_strategy.remove_unnecessary_lock = False
build_strategy.memory_optimize = False
if num_threads > 1:
build_strategy.reduce_strategy = F.BuildStrategy.ReduceStrategy.Reduce
val_compiled_prog = F.compiler.CompiledProgram(
val_program).with_data_parallel(
build_strategy=build_strategy,
exec_strategy=exec_strategy)
num_nodes = int(np.load(os.path.join(config.graph_work_path, "num_nodes.npy")))
predict_data = PredictData(num_nodes)
predict_iter = NodeClassificationGenerator(
graph_wrappers=model.graph_wrappers,
batch_size=config.infer_batch_size,
data=predict_data,
samples=config.samples,
num_workers=config.sample_workers,
feed_name_list=[var.name for var in model.feed_list],
use_pyreader=config.use_pyreader,
phase="predict",
graph_data_path=config.graph_work_path,
shuffle=False)
if config.learner_type == "cpu":
model.data_loader.decorate_batch_generator(
predict_iter, places=F.cpu_places())
elif config.learner_type == "gpu":
gpu_id = int(os.getenv("FLAGS_selected_gpus", "0"))
place = F.CUDAPlace(gpu_id)
model.data_loader.decorate_batch_generator(
predict_iter, places=place)
else:
raise ValueError
run_predict(model.data_loader,
program=val_compiled_prog,
exe=exe,
model_dict=model,
args=config)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='main')
parser.add_argument("--conf", type=str, default="./config.yaml")
args = parser.parse_args()
config = edict(yaml.load(open(args.conf), Loader=yaml.FullLoader))
print(config)
main(config)