yaml:
yet another markup language,另一种标记语言,专门用来写配置文件,类似于json格式。
语法
- 大小写敏感
- 使用缩进表示层级关系
- 缩进时不允许使用Tab键,只允许使用空格
- 缩进的空格数目不重要,只要相同层级的元素左侧对齐即可
- # 表示注释,从这个字符一直到行尾,都会被解析器忽略
数据结构:
- 对象:键值对的集合,又称为映射/哈希/字典
- 数组:一组按次序排列的值,又称为序列/列表
- 纯量:单个的、不可再分的值。字符串、布尔值、整数、浮点数、null、时间、日期
序列:
1.yaml里面写一个数组,前面加一个‘-’符号,如下
- admin1: 123456
- admin2: 111111
- admin3: 222222
对应python里面的list
[{'admin1': 123456},
{'admin2': 111111},
{'admin3': 222222}]
注意:数字读出来的是int或float类型
纯量:
在yaml文件写入以下内容:
n1: 12.30
n2: true
n3: false
n4: ~
time1: 2018-04-18t21:59:43.10+08:00
date1: 2018-04-18
n6: !!str 123
n7: !!str true
python读取结果:
{'n1': 12.3,
'n2': True,
'n3': False,
'n4': None,
'time1': datetime.datetime(2018, 4, 18, 13, 59, 43, 100000),
'date1': datetime.date(2018, 4, 18),
'n6': '123',
'n7': 'true'}
混合使用:
1.list嵌套dict,在yaml里面写入如下内容:
- user: admin1
psw: '123456'
- user: admin2
psw: '111111'
- user: admin3
psw: '222222'
用python读取出来的结果:
[{'user': 'admin1', 'psw': '123456'},
{'user': 'admin2', 'psw': '111111'},
{'user': 'admin3', 'psw': '222222'}]
2.dict嵌套list,在yaml里面写入如下内容:
nub1:
- admin1
- '123456'
nb2:
- admin2
- '111111'
nb3:
- admin3
- '222222'
用python读取出来的结果:
{'nub1': ['admin1', '123456'],
'nb2': ['admin2', '111111'],
'nb3': ['admin3', '222222']
读取yaml格式文件:
with open('./config.yml') as file_config:
config=yaml.load(file_config)
feature_names=config['model_params']['feature_names']
#以下是config配置内容例子
data_params:
path_result: ../data/lstm_crf/result.txt
path_test: ../data/lstm_crf/test.txt
path_train: ../data/lstm_crf/train.txt
sep: table
voc_params:
char:
min_count: 1
path: null
f1:
min_count: 1
path: ../data/lstm_crf/train_test_word2ids.pkl
f2:
min_count: 1
path: null
label:
min_count: 0
path: ../data/lstm_crf/train_test_tag2ids.pkl
model: lstm_crf
model_params:
batch_size: 128
bilstm_params:
num_layers: 1
num_units: 256
rnn_dropout: 0.2
clip: 10
conv_dropout: 0.2
conv_filter_len_list:
- 1
- 2
- 3
- 4
- 5
conv_filter_size_list:
- 8
- 8
- 8
- 8
- 8
dev_size: 0.1
dropout_rate: 0.3
embed_params:
char:
dropout_rate: null
path: null
path_pre_train: null
shape:
- null
- null
f1:
dropout_rate: 0.3
path: ../data/lstm_crf/f1_embed.mat.pkl
path_pre_train: /E/home/mayajun/PycharmProjects/daguan_lstm_crf/data/word2vec.txt
shape:
- 4685
- 256
f2:
dropout_rate: 0.3
path: null
path_pre_train: null
shape:
- null
- null
feature_names:
- f1
l2_rate: 0.001
learning_rate: 5.0e-05
max_patience: 5
nb_classes: 4
nb_epoch: 1000
path_model: ../model/lstm_crf/model
rnn_unit: lstm
sequence_len_pt: 95
sequence_length: 128
use_char_feature: false
use_crf: true
word_len_pt: 95
word_length: 11
unknow_word: unk