Skip to content

Commit 340402e

Browse files
author
Igor Rukhovich
committed
Added abalone + letters, updated all GB configs
1 parent dcfc5b9 commit 340402e

File tree

10 files changed

+489
-300
lines changed

10 files changed

+489
-300
lines changed

‎configs/lgbm_mb_cpu_config.json

Lines changed: 78 additions & 66 deletions
Original file line numberDiff line numberDiff line change
@@ -1,108 +1,120 @@
11
{
22
"common": {
3-
"lib": ["modelbuilders"],
4-
"data-format": ["pandas"],
5-
"data-order": ["F"],
6-
"dtype": ["float32"]
3+
"lib": "modelbuilders",
4+
"data-format": "pandas",
5+
"data-order": "F",
6+
"dtype": "float32",
7+
"algorithm": "lgbm_mb"
78
},
89
"cases": [
910
{
10-
"algorithm": "lgbm_mb",
1111
"dataset": [
1212
{
13-
"source": "csv",
14-
"name": "mortgage1Q",
13+
"source": "npy",
14+
"name": "airline-ohe",
1515
"training":
1616
{
17-
"x": "data/mortgage_x.csv",
18-
"y": "data/mortgage_y.csv"
17+
"x": "data/airline-ohe_x_train.npy",
18+
"y": "data/airline-ohe_y_train.npy"
19+
},
20+
"testing":
21+
{
22+
"x": "data/airline-ohe_x_test.npy",
23+
"y": "data/airline-ohe_y_test.npy"
1924
}
2025
}
2126
],
22-
"n-estimators": [100],
23-
"objective": ["regression"],
24-
"max-depth": [8],
25-
"scale-pos-weight": [2],
26-
"learning-rate": [0.1],
27-
"subsample": [1],
28-
"reg-alpha": [0.9],
29-
"reg-lambda": [1],
30-
"min-child-weight": [0],
31-
"max-leaves": [256]
27+
"reg-alpha": 0.9,
28+
"max-bin": 256,
29+
"scale-pos-weight": 2,
30+
"learning-rate": 0.1,
31+
"subsample": 1,
32+
"reg-lambda": 1,
33+
"min-child-weight": 0,
34+
"max-depth": 8,
35+
"max-leaves": 256,
36+
"n-estimators": 1000,
37+
"objective": "binary"
3238
},
3339
{
34-
"algorithm": "lgbm_mb",
3540
"dataset": [
3641
{
37-
"source": "csv",
38-
"name": "airline-ohe",
42+
"source": "npy",
43+
"name": "higgs1m",
3944
"training":
4045
{
41-
"x": "data/airline-ohe_x_train.csv",
42-
"y": "data/airline-ohe_y_train.csv"
46+
"x": "data/higgs1m_x_train.npy",
47+
"y": "data/higgs1m_y_train.npy"
48+
},
49+
"testing":
50+
{
51+
"x": "data/higgs1m_x_test.npy",
52+
"y": "data/higgs1m_y_test.npy"
4353
}
4454
}
4555
],
46-
"reg-alpha": [0.9],
47-
"max-bin": [256],
48-
"scale-pos-weight": [2],
49-
"learning-rate": [0.1],
50-
"subsample": [1],
51-
"reg-lambda": [1],
52-
"min-child-weight": [0],
53-
"max-depth": [8],
54-
"max-leaves": [256],
55-
"n-estimators": [1000],
56-
"objective": ["binary"]
56+
"reg-alpha": 0.9,
57+
"max-bin": 256,
58+
"scale-pos-weight": 2,
59+
"learning-rate": 0.1,
60+
"subsample": 1,
61+
"reg-lambda": 1,
62+
"min-child-weight": 0,
63+
"max-depth": 8,
64+
"max-leaves": 256,
65+
"n-estimators": 1000,
66+
"objective": "binary"
5767
},
5868
{
59-
"algorithm": "lgbm_mb",
6069
"dataset": [
6170
{
62-
"source": "csv",
63-
"name": "higgs1m",
71+
"source": "csv",
72+
"name": "mortgage1Q",
6473
"training":
6574
{
66-
"x": "data/higgs1m_x_train.csv",
67-
"y": "data/higgs1m_y_train.csv"
75+
"x": "data/mortgage_x.csv",
76+
"y": "data/mortgage_y.csv"
6877
}
6978
}
7079
],
71-
"reg-alpha": [0.9],
72-
"max-bin": [256],
73-
"scale-pos-weight": [2],
74-
"learning-rate": [0.1],
75-
"subsample": [1],
76-
"reg-lambda": [1],
77-
"min-child-weight": [0],
78-
"max-depth": [8],
79-
"max-leaves": [256],
80-
"n-estimators": [1000],
81-
"objective": ["binary"]
80+
"n-estimators": 100,
81+
"objective": "regression",
82+
"max-depth": 8,
83+
"scale-pos-weight": 2,
84+
"learning-rate": 0.1,
85+
"subsample": 1,
86+
"reg-alpha": 0.9,
87+
"reg-lambda": 1,
88+
"min-child-weight": 0,
89+
"max-leaves": 256
8290
},
8391
{
84-
"algorithm": "lgbm_mb",
8592
"dataset": [
8693
{
87-
"source": "csv",
88-
"name": "msrank",
94+
"source": "npy",
95+
"name": "msrank",
8996
"training":
9097
{
91-
"x": "data/mlsr_x_train.csv",
92-
"y": "data/mlsr_y_train.csv"
98+
"x": "data/msrank_x_train.npy",
99+
"y": "data/msrank_y_train.npy"
100+
},
101+
"testing":
102+
{
103+
"x": "data/msrank_x_test.npy",
104+
"y": "data/msrank_y_test.npy"
93105
}
94106
}
95107
],
96-
"max-bin": [256],
97-
"learning-rate": [0.3],
98-
"subsample": [1],
99-
"reg-lambda": [2],
100-
"min-child-weight": [1],
101-
"min-split-gain": [0.1],
102-
"max-depth": [8],
103-
"max-leaves": [256],
104-
"n-estimators": [200],
105-
"objective": ["multiclass"]
108+
"max-bin": 256,
109+
"learning-rate": 0.3,
110+
"subsample": 1,
111+
"reg-lambda": 2,
112+
"min-child-weight": 1,
113+
"min-split-loss": 0.1,
114+
"max-depth": 8,
115+
"max-leaves": 256,
116+
"n-estimators": 200,
117+
"objective": "multiclass"
106118
}
107119
]
108120
}

‎configs/xgb_cpu_config.json

Lines changed: 99 additions & 56 deletions
Original file line numberDiff line numberDiff line change
@@ -4,77 +4,32 @@
44
"data-format": "pandas",
55
"data-order": "F",
66
"dtype": "float32",
7-
"count-dmatrix":"",
87
"algorithm": "gbt",
98
"tree-method": "hist",
10-
"num-threads": 56
9+
"count-dmatrix":""
1110
},
1211
"cases": [
1312
{
1413
"dataset": [
1514
{
16-
"source": "csv",
17-
"name": "plasticc",
15+
"source": "npy",
16+
"name": "abalone",
1817
"training":
1918
{
20-
"x": "data/plasticc_x_train.csv",
21-
"y": "data/plasticc_y_train.csv"
19+
"x": "data/abalone_x_train.npy",
20+
"y": "data/abalone_y_train.npy"
2221
},
2322
"testing":
2423
{
25-
"x": "data/plasticc_x_test.csv",
26-
"y": "data/plasticc_y_test.csv"
27-
}
28-
}
29-
],
30-
"n-estimators": 60,
31-
"objective": "multi:softprob",
32-
"max-depth": 7,
33-
"subsample": 0.7,
34-
"colsample-bytree": 0.7
35-
},
36-
{
37-
"dataset": [
38-
{
39-
"source": "csv",
40-
"name": "santander",
41-
"training":
42-
{
43-
"x": "data/santander_x_train.csv",
44-
"y": "data/santander_y_train.csv"
45-
}
46-
}
47-
],
48-
"n-estimators": 10000,
49-
"objective": "binary:logistic",
50-
"max-depth": 1,
51-
"subsample": 0.5,
52-
"eta": 0.1,
53-
"colsample-bytree": 0.05,
54-
"single-precision-histogram": ""
55-
},
56-
{
57-
"dataset": [
58-
{
59-
"source": "csv",
60-
"name": "mortgage1Q",
61-
"training":
62-
{
63-
"x": "data/mortgage_x.csv",
64-
"y": "data/mortgage_y.csv"
24+
"x": "data/abalone_x_test.npy",
25+
"y": "data/abalone_y_test.npy"
6526
}
6627
}
6728
],
68-
"n-estimators": 100,
69-
"objective": "reg:squarederror",
70-
"max-depth": 8,
71-
"scale-pos-weight": 2,
72-
"learning-rate": 0.1,
73-
"subsample": 1,
74-
"reg-alpha": 0.9,
75-
"reg-lambda": 1,
76-
"min-child-weight": 0,
77-
"max-leaves": 256
29+
"learning-rate": 0.03,
30+
"max-depth": 6,
31+
"n-estimators": 1000,
32+
"objective": "reg:squarederror"
7833
},
7934
{
8035
"dataset": [
@@ -136,6 +91,51 @@
13691
"enable-experimental-json-serialization": "False",
13792
"inplace-predict": ""
13893
},
94+
{
95+
"dataset": [
96+
{
97+
"source": "npy",
98+
"name": "letters",
99+
"training":
100+
{
101+
"x": "data/letters_x_train.npy",
102+
"y": "data/letters_y_train.npy"
103+
},
104+
"testing":
105+
{
106+
"x": "data/letters_x_test.npy",
107+
"y": "data/letters_y_test.npy"
108+
}
109+
}
110+
],
111+
"learning-rate": 0.03,
112+
"max-depth": 6,
113+
"n-estimators": 1000,
114+
"objective": "multi:softprob"
115+
},
116+
{
117+
"dataset": [
118+
{
119+
"source": "csv",
120+
"name": "mortgage1Q",
121+
"training":
122+
{
123+
"x": "data/mortgage_x.csv",
124+
"y": "data/mortgage_y.csv"
125+
}
126+
}
127+
],
128+
"n-estimators": 100,
129+
"objective": "reg:squarederror",
130+
"max-depth": 8,
131+
"scale-pos-weight": 2,
132+
"learning-rate": 0.1,
133+
"subsample": 1,
134+
"reg-alpha": 0.9,
135+
"reg-lambda": 1,
136+
"min-child-weight": 0,
137+
"max-leaves": 256
138+
},
139139
{
140140
"dataset": [
141141
{
@@ -163,6 +163,49 @@
163163
"n-estimators": 200,
164164
"objective": "multi:softprob",
165165
"single-precision-histogram": ""
166+
},
167+
{
168+
"dataset": [
169+
{
170+
"source": "csv",
171+
"name": "plasticc",
172+
"training":
173+
{
174+
"x": "data/plasticc_x_train.csv",
175+
"y": "data/plasticc_y_train.csv"
176+
},
177+
"testing":
178+
{
179+
"x": "data/plasticc_x_test.csv",
180+
"y": "data/plasticc_y_test.csv"
181+
}
182+
}
183+
],
184+
"n-estimators": 60,
185+
"objective": "multi:softprob",
186+
"max-depth": 7,
187+
"subsample": 0.7,
188+
"colsample-bytree": 0.7
189+
},
190+
{
191+
"dataset": [
192+
{
193+
"source": "csv",
194+
"name": "santander",
195+
"training":
196+
{
197+
"x": "data/santander_x_train.csv",
198+
"y": "data/santander_y_train.csv"
199+
}
200+
}
201+
],
202+
"n-estimators": 10000,
203+
"objective": "binary:logistic",
204+
"max-depth": 1,
205+
"subsample": 0.5,
206+
"eta": 0.1,
207+
"colsample-bytree": 0.05,
208+
"single-precision-histogram": ""
166209
}
167210
]
168211
}

0 commit comments

Comments
 (0)