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
|
import operator_benchmark as op_bench
"""
Configs shared by multiple benchmarks
"""
def remove_cuda(config_list):
cuda_config = {"device": "cuda"}
return [config for config in config_list if cuda_config not in config]
# Configs for conv-1d ops
conv_1d_configs_short = op_bench.config_list(
attr_names=["IC", "OC", "kernel", "stride", "N", "L"],
attrs=[
[128, 256, 3, 1, 1, 64],
[256, 256, 3, 2, 4, 64],
],
cross_product_configs={
"device": ["cpu", "cuda"],
},
tags=["short"],
)
conv_1d_configs_long = op_bench.cross_product_configs(
IC=[128, 512],
OC=[128, 512],
kernel=[3],
stride=[1, 2],
N=[8],
L=[128],
device=["cpu", "cuda"],
tags=["long"],
)
convtranspose_1d_configs_short = op_bench.config_list(
attr_names=["IC", "OC", "kernel", "stride", "N", "L"],
attrs=[
[2016, 1026, 1024, 256, 1, 224],
],
cross_product_configs={
"device": ["cpu", "cuda"],
},
tags=["short"],
)
# Configs for Conv2d and ConvTranspose1d
conv_2d_configs_short = op_bench.config_list(
attr_names=[
"IC",
"OC",
"kernel",
"stride",
"N",
"H",
"W",
"G",
"pad",
],
attrs=[
[256, 256, 3, 1, 1, 16, 16, 1, 0],
],
cross_product_configs={
"device": ["cpu", "cuda"],
},
tags=["short"],
)
conv_2d_configs_long = op_bench.cross_product_configs(
IC=[128, 256],
OC=[128, 256],
kernel=[3],
stride=[1, 2],
N=[4],
H=[32],
W=[32],
G=[1],
pad=[0],
device=["cpu", "cuda"],
tags=["long"],
)
# Configs for Conv2dPointwise
conv_2d_pw_configs_short = op_bench.config_list(
attr_names=[
"IC",
"OC",
"stride",
"N",
"H",
"W",
"G",
"pad",
],
attrs=[
[256, 256, 1, 1, 16, 16, 1, 0],
],
cross_product_configs={
"device": ["cpu", "cuda"],
},
tags=["short"],
)
conv_2d_pw_configs_long = op_bench.cross_product_configs(
IC=[128, 256],
OC=[128, 256],
stride=[1, 2],
N=[4],
H=[32],
W=[32],
G=[1],
pad=[0],
device=["cpu", "cuda"],
tags=["long"],
)
# Configs for Conv3d and ConvTranspose3d
conv_3d_configs_short = op_bench.config_list(
attr_names=["IC", "OC", "kernel", "stride", "N", "D", "H", "W"],
attrs=[
[64, 64, 3, 1, 8, 4, 16, 16],
],
cross_product_configs={
"device": ["cpu", "cuda"],
},
tags=["short"],
)
linear_configs_short = op_bench.config_list(
attr_names=["N", "IN", "OUT"],
attrs=[
[1, 1, 1],
[4, 256, 128],
[16, 512, 256],
],
cross_product_configs={
"device": ["cpu", "cuda"],
},
tags=["short"],
)
linear_configs_long = op_bench.cross_product_configs(
N=[32, 64], IN=[128, 512], OUT=[64, 128], device=["cpu", "cuda"], tags=["long"]
)
embeddingbag_short_configs = op_bench.cross_product_configs(
embeddingbags=[10, 120, 1000, 2300],
dim=[64],
mode=["sum"],
input_size=[8, 16, 64],
offset=[0],
sparse=[True, False],
include_last_offset=[True, False],
device=["cpu"],
tags=["short"],
)
embedding_short_configs = op_bench.cross_product_configs(
num_embeddings=[10, 120, 1000, 2300],
embedding_dim=[64],
input_size=[8, 16, 64],
device=["cpu"],
tags=["short"],
)
|
