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
|
// META: title=validation tests for pooling and reduction operators keep dimensions
// META: global=window,dedicatedworker
// META: script=../resources/utils.js
// META: script=../resources/utils_validation.js
// META: timeout=long
'use strict';
// This is used to reproduce an issue(crbug.com/331841268) of averagePool2d in
// ResNetV2 50 model.
// [input]
// |
// [globalAveragePool]
// |
// [conv2d]
// |
// [reshape]
// |
// [output]
promise_test(async t => {
const avgPool2dInputShape = [1, 7, 7, 2048];
const avgPool2dInput = builder.input(
`avgPool2dInput`, {dataType: 'float32', dimensions: avgPool2dInputShape});
const avgPool2dOutput =
builder.averagePool2d(avgPool2dInput, {layout: 'nhwc'});
const conv2dFilterShape = [1001, 1, 1, 2048];
const conv2dFilter = builder.constant(
{dataType: 'float32', dimensions: conv2dFilterShape},
new Float32Array(sizeOfShape(conv2dFilterShape)).fill(1));
const conv2dBias = builder.constant(
{dataType: 'float32', dimensions: [1001]},
new Float32Array(1001).fill(0.01));
const conv2dOutput = builder.conv2d(avgPool2dOutput, conv2dFilter, {
inputLayout: 'nhwc',
filterLayout: 'ohwi',
padding: [0, 0, 0, 0],
bias: conv2dBias
});
const newShape = [1, 1001];
const reshapeOutput = builder.reshape(conv2dOutput, newShape);
assert_equals(reshapeOutput.dataType(), avgPool2dInput.dataType());
assert_array_equals(reshapeOutput.shape(), newShape);
const graph = await builder.build({reshapeOutput});
const result = await context.compute(
graph, {
'avgPool2dInput':
new Float32Array(sizeOfShape(avgPool2dInputShape)).fill(0.1)
},
{'reshapeOutput': new Float32Array(1001)});
}, 'Test global average pool operator\'s output shape for ResNetV2 50 model.');
// This is used to reproduce an issue(crbug.com/331841268) of reduceMean in
// ResNetV2 50 model.
// [input]
// |
// [reduceMean]
// |
// [conv2d]
// |
// [reshape]
// |
// [output]
promise_test(async t => {
const reduceMeanInputShape = [1, 7, 7, 2048];
const reduceMeanInput = builder.input(
`reduceMeanInput`,
{dataType: 'float32', dimensions: reduceMeanInputShape});
const reduceMeanOutput =
builder.reduceMean(reduceMeanInput, {axes: [1, 2], keepDimensions: true});
const conv2dFilterShape = [1001, 1, 1, 2048];
const conv2dFilter = builder.constant(
{dataType: 'float32', dimensions: conv2dFilterShape},
new Float32Array(sizeOfShape(conv2dFilterShape)).fill(1));
const conv2dBias = builder.constant(
{dataType: 'float32', dimensions: [1001]},
new Float32Array(1001).fill(0.01));
const conv2dOutput = builder.conv2d(reduceMeanOutput, conv2dFilter, {
inputLayout: 'nhwc',
filterLayout: 'ohwi',
padding: [0, 0, 0, 0],
bias: conv2dBias
});
const newShape = [1, 1001];
const reshapeOutput = builder.reshape(conv2dOutput, newShape);
assert_equals(reshapeOutput.dataType(), reduceMeanInput.dataType());
assert_array_equals(reshapeOutput.shape(), newShape);
const graph = await builder.build({reshapeOutput});
const result = await context.compute(
graph, {
'reduceMeanInput':
new Float32Array(sizeOfShape(reduceMeanInputShape)).fill(0.1)
},
{'reshapeOutput': new Float32Array(1001)});
}, 'Test reduceMean operator\'s output shape for ResNetV2 50 model.');
|