This repository was archived by the owner on Jun 22, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 8
/
Copy pathpipeline_config.py
178 lines (164 loc) · 8.75 KB
/
pipeline_config.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
from itertools import product
import os
from attrdict import AttrDict
from deepsense import neptune
from nltk.corpus import stopwords
import numpy as np
from utils import read_params, safe_eval
ctx = neptune.Context()
params = read_params(ctx)
FEATURE_COLUMNS = ['user_id',
'region', 'city',
'parent_category_name', 'category_name',
'param_1', 'param_2', 'param_3',
'title', 'description',
'price',
'item_seq_number',
'activation_date',
'user_type',
'image',
'image_top_1']
CATEGORICAL_COLUMNS = ['region', 'city',
'parent_category_name', 'category_name',
'param_1', 'param_2', 'param_3',
'user_type', 'image_top_1']
NUMERICAL_COLUMNS = ['price', 'item_seq_number']
TEXT_COLUMNS = ['title', 'description']
IMAGE_COLUMNS = ['image']
TARGET_COLUMNS = ['deal_probability']
CV_COLUMN = ['user_id']
TIMESTAMP_COLUMNS = ['activation_date']
ITEM_ID_COLUMN = ['item_id']
USER_ID_COLUMN = ['user_id']
FEATURES_TO_TRANSLATE = ['category_name', 'city', 'description', 'param_1', 'param_2', 'param_3',
'parent_category_name', 'region', 'title']
DEV_SAMPLE_SIZE = int(10e1)
COLUMN_TYPES = {'train': {'price': 'float64',
'item_seq_number': 'uint32',
'image_top_1': 'float64',
'deal_probability': 'float32',
},
'inference': {'price': 'float64',
'item_seq_number': 'uint32',
'image_top_1': 'float64',
}
}
AGGREGATION_RECIPIES = []
for time, agg in product([['activation_date_month'], ['activation_date_day', 'activation_date_month']],
['mean', 'size', 'var', 'min', 'max']):
for group in [['user_id'] + time,
['category_name'] + time,
['parent_category_name'] + time,
['param_1'] + time,
['param_2'] + time,
['param_3'] + time,
['user_id', 'category_name'] + time,
['user_id', 'parent_category_name'] + time,
['user_id', 'param_1'] + time,
['user_id', 'param_2'] + time,
['user_id', 'param_3'] + time,
['user_type'] + time,
['user_type', 'category_name'] + time,
['user_type', 'parent_category_name'] + time,
['user_type', 'param_1'] + time,
['user_type', 'param_2'] + time,
['user_type', 'param_3'] + time,
['user_type', 'city'] + time,
['user_type', 'city', 'category_name'] + time,
['user_type', 'city', 'parent_category_name'] + time,
['user_type', 'region'] + time,
['user_type', 'region', 'category_name'] + time,
['user_type', 'region', 'parent_category_name'] + time,
['user_type', 'region', 'param_1'] + time,
['user_type', 'region', 'param_2'] + time,
['user_type', 'region', 'param_3'] + time,
]:
AGGREGATION_RECIPIES.append({'groupby': group, 'select': 'price', 'agg': agg})
SOLUTION_CONFIG = AttrDict({
'env': {'cache_dirpath': params.experiment_dir
},
'random_search': {'light_gbm': {'n_runs': safe_eval(params.lgbm_random_search_runs),
'callbacks': {'neptune_monitor': {'name': 'light_gbm'
},
'save_results': {
'filepath': os.path.join(params.experiment_dir,
'random_search_light_gbm.pkl')
}
}
}
},
'input_missing': {'text_columns': (TEXT_COLUMNS, '<this_is_missing_value>'),
'categorical_columns': (CATEGORICAL_COLUMNS, '-9999'),
'numerical_columns': (NUMERICAL_COLUMNS, 0),
'timestamp_columns': (TIMESTAMP_COLUMNS, '2017-03-15')
},
'date_features': {'date_column': TIMESTAMP_COLUMNS[0]},
'is_missing': {'columns': FEATURE_COLUMNS},
'categorical_encoder': {},
'groupby_aggregation': {'groupby_aggregations': AGGREGATION_RECIPIES
},
'target_encoder': {'n_splits': safe_eval(params.target_encoder__n_splits),
},
'text_features': {'cols': ['description', 'title']},
'word_overlap': {'overlap_cols': [('description', 'title'),
('description', 'parent_category_name'),
('description', 'category_name'),
('description', 'param_1'),
('description', 'param_2'),
('description', 'param_3'),
('title', 'parent_category_name'),
('title', 'category_name'),
('title', 'param_1'),
('title', 'param_2'),
('title', 'param_3'),
]},
'tfidf': {'cols_params': [('description', {'ngram_range': (1, 2),
'max_features': 16000,
"stop_words": set(stopwords.words('english')),
"analyzer": 'word',
"token_pattern": r'\w{1,}',
"sublinear_tf": True,
"dtype": np.float32,
"norm": 'l2',
"smooth_idf": False
}),
('title', {'ngram_range': (1, 2),
'max_features': 8000,
"stop_words": set(stopwords.words('english')),
"analyzer": 'word',
"token_pattern": r'\w{1,}',
"sublinear_tf": True,
"dtype": np.float32,
"norm": 'l2',
"smooth_idf": False
})]},
'image_stats': {'cols': IMAGE_COLUMNS,
'img_dir_train': params.train_image_dir,
'img_dir_test': params.test_image_dir,
'log_features': True,
'n_jobs': params.num_workers},
'light_gbm': {'boosting_type': safe_eval(params.lgbm__boosting_type),
'objective': safe_eval(params.lgbm__objective),
'metric': safe_eval(params.lgbm__metric),
'learning_rate': safe_eval(params.lgbm__learning_rate),
'max_depth': safe_eval(params.lgbm__max_depth),
'subsample': safe_eval(params.lgbm__subsample),
'colsample_bytree': safe_eval(params.lgbm__colsample_bytree),
'min_child_weight': safe_eval(params.lgbm__min_child_weight),
'reg_lambda': safe_eval(params.lgbm__reg_lambda),
'reg_alpha': safe_eval(params.lgbm__reg_alpha),
'subsample_freq': safe_eval(params.lgbm__subsample_freq),
'max_bin': safe_eval(params.lgbm__max_bin),
'min_child_samples': safe_eval(params.lgbm__min_child_samples),
'num_leaves': safe_eval(params.lgbm__num_leaves),
'top_rate': safe_eval(params.lgbm__top_rate),
'other_rate': safe_eval(params.lgbm__other_rate),
'nthread': safe_eval(params.num_workers),
'number_boosting_rounds': safe_eval(params.lgbm__number_boosting_rounds),
'early_stopping_rounds': safe_eval(params.lgbm__early_stopping_rounds),
'scale_pos_weight': safe_eval(params.lgbm__scale_pos_weight),
'verbose': safe_eval(params.verbose)
},
'clipper': {'min_val': 0,
'max_val': 1}
})