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
|
import torch
import operator_benchmark as op_bench
qcomparators_configs = op_bench.cross_product_configs(
N=(8, 64),
dtype=(torch.quint8, torch.qint8, torch.qint32),
contig=(False, True),
other_scalar=(False, True),
out_variant=(False, True),
tags=('short',)
)
qcomparators_ops = op_bench.op_list(
attrs=(
('eq', torch.eq),
('ne', torch.ne),
('lt', torch.lt),
('gt', torch.gt),
('le', torch.le),
('ge', torch.ge),
),
attr_names=('op_name', 'op_func'),
)
class QComparatorBenchmark(op_bench.TorchBenchmarkBase):
def init(self, N, dtype, contig, other_scalar, out_variant, op_func):
# TODO: Consider more diverse shapes
f_input = (torch.rand(N, N) - 0.5) * 256
scale = 1.0
zero_point = 0
q_input_a = torch.quantize_per_tensor(f_input, scale=scale,
zero_point=zero_point,
dtype=dtype)
q_input_b = q_input_a.clone()
if not contig:
permute_dims = list(range(f_input.ndim))[::-1]
q_input_a = q_input_a.permute(permute_dims)
self.qop = op_func
self.inputs = {
"q_input_a": q_input_a,
"q_input_b": q_input_b,
"out_variant": out_variant,
"other_scalar": other_scalar,
}
def forward(self, q_input_a, q_input_b, out_variant: bool, other_scalar: bool):
if out_variant:
if other_scalar:
return self.qop(q_input_a, 42, out=torch.tensor(True, dtype=torch.bool))
else:
return self.qop(q_input_a, q_input_b, out=torch.tensor(True, dtype=torch.bool))
else:
if other_scalar:
return self.qop(q_input_a, 42)
else:
return self.qop(q_input_a, q_input_b)
op_bench.generate_pt_tests_from_op_list(qcomparators_ops,
qcomparators_configs,
QComparatorBenchmark)
if __name__ == '__main__':
op_bench.benchmark_runner.main()
|
