-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrun.py
119 lines (100 loc) · 4.28 KB
/
run.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
'''
* @author Waldinsamkeit
* @email Zenglz_pro@163.com
* @create date 2020-10-29 19:02:44
* @desc
'''
from typing import Any,Dict
from dataloader import DataSetDir, DataSet, Dataloader, DataItemSet,select_sampler
from wrapper import ModelWrapper
from model import Embedding_layer, Scorer, SetinstanceClassifier
from evaluate import select_evaluate_func
import config
from config import TrainingConfig,OperateConfig,DataConfig,ModelConfig
from log import logger
from utils import set_random_seed
SEED = 2020
def test_clustertask(operateconfig:Dict,dataconfig:Dict, trainingconfig:Dict, modelconfig:Dict):
dir_path = dataconfig['data_dir_path']
if not dir_path:
raise KeyError
datasetdir = DataSetDir(dir_path,word_emb_select=dataconfig['word_emb_select'])
# combine model
embedding_layer = Embedding_layer.from_pretrained(datasetdir.embedding_vec)
embedding_layer.freeze_parameters()
scorer = Scorer(
embedding_layer,
modelconfig['embed_trans_hidden_size'],
modelconfig['post_trans_hidden_size'],
modelconfig['dropout']
)
model = SetinstanceClassifier(
scorer=scorer,
name=modelconfig['name'],
version=modelconfig['version']
)
wrapper = ModelWrapper(model,trainingconfig)
if operateconfig['resume']:
wrapper.load_check_point()
# continue to trainning
if operateconfig['train']:
train_datasetitem = DataItemSet(
dataset=datasetdir.train_dataset,
sampler = select_sampler(dataconfig['sample_strategy']),
negative_sample_size = dataconfig['negative_sample_size']
)
dev_datasetitem = DataItemSet(
dataset=datasetdir.dev_dataset,
sampler = select_sampler(dataconfig['sample_strategy']),
negative_sample_size = dataconfig['test_negative_sample_size']
)
train_dataloader = Dataloader(
dataitems=train_datasetitem,
word2id=datasetdir.word2id,
batch_size=trainingconfig['batch_size']
)
dev_dataloader = Dataloader(
dataitems=dev_datasetitem,
word2id=datasetdir.word2id,
batch_size=trainingconfig['batch_size']
)
wrapper.train(train_dataloader=train_dataloader,dev_dataloader=dev_dataloader)
if operateconfig['test']:
test_datasetitem = DataItemSet(
dataset=datasetdir.test_dataset,
sampler = select_sampler(dataconfig['sample_strategy']),
negative_sample_size = dataconfig['test_negative_sample_size']
)
test_dataloader = Dataloader(
dataitems=test_datasetitem,
word2id=datasetdir.word2id,
batch_size=trainingconfig['batch_size']
)
wrapper.test(test_dataloader=test_dataloader)
if operateconfig['predict']:
func_list = select_evaluate_func(operateconfig['eval_function'])
# import pdb;pdb.set_trace()
pred_word_set = wrapper.cluster_predict(
dataset=datasetdir.test_dataset,
word2id=datasetdir.word2id,
outputfile=trainingconfig['result_out_dir'].joinpath(datasetdir.name+'_result.txt')
)
# import pdb;pdb.set_trace()
ans = wrapper.evaluate(datasetdir.test_dataset, pred_word_set,function_list=func_list)
logger.info("{} DataSet Cluster Prediction".format(datasetdir.train_dataset.name))
for name,f in ans:
logger.info("{} : {:.2f}".format(name,f))
def NYT():
DataConfig['data_dir_path'] = config.NYT_DIR_PATH
test_clustertask(OperateConfig,DataConfig,TrainingConfig,ModelConfig)
def PubMed():
DataConfig['data_dir_path'] = config.PubMed_DIR_PATH
test_clustertask(OperateConfig,DataConfig,TrainingConfig,ModelConfig)
def Wiki():
DataConfig['data_dir_path'] = config.Wiki_DIR_PATH
test_clustertask(OperateConfig,DataConfig,TrainingConfig,ModelConfig)
if __name__ == '__main__':
set_random_seed(seed=SEED)
NYT()
PubMed()
Wiki()