Source code
Revision control
Copy as Markdown
Other Tools
// META: title=validation tests for WebNN API instanceNormalization operation
// META: global=window
// META: variant=?cpu
// META: variant=?gpu
// META: variant=?npu
// META: script=../resources/utils_validation.js
'use strict';
const kExampleInputDescriptor = {
dataType: 'float32',
shape: [2, 2, 2, 2]
};
// 1D tensor descriptor which may be used for `scale`, or `bias` inputs.
const kExample1DTensorDescriptor = {
dataType: 'float32',
shape: [2]
};
multi_builder_test(async (t, builder, otherBuilder) => {
const inputFromOtherBuilder =
otherBuilder.input('input', kExampleInputDescriptor);
assert_throws_js(
TypeError, () => builder.instanceNormalization(inputFromOtherBuilder));
}, '[instanceNormalization] throw if input is from another builder');
multi_builder_test(async (t, builder, otherBuilder) => {
const scaleFromOtherBuilder =
otherBuilder.input('scale', kExample1DTensorDescriptor);
const options = {scale: scaleFromOtherBuilder};
const input = builder.input('input', kExampleInputDescriptor);
assert_throws_js(
TypeError, () => builder.instanceNormalization(input, options));
}, '[instanceNormalization] throw if scale option is from another builder');
multi_builder_test(async (t, builder, otherBuilder) => {
const biasFromOtherBuilder =
otherBuilder.input('bias', kExample1DTensorDescriptor);
const options = {bias: biasFromOtherBuilder};
const input = builder.input('input', kExampleInputDescriptor);
assert_throws_js(
TypeError, () => builder.instanceNormalization(input, options));
}, '[instanceNormalization] throw if bias option is from another builder');
const label = 'instance_normalization';
const tests = [
{
name: '[instanceNormalization] Test with default options for 4-D input.',
input: {dataType: 'float32', shape: [1, 2, 3, 4]},
output: {dataType: 'float32', shape: [1, 2, 3, 4]}
},
{
name:
'[instanceNormalization] Test with scale, bias and default epsilon value.',
input: {dataType: 'float32', shape: [1, 2, 3, 4]},
options: {
scale: {dataType: 'float32', shape: [2]},
bias: {dataType: 'float32', shape: [2]},
epsilon: 1e-5,
},
output: {dataType: 'float32', shape: [1, 2, 3, 4]}
},
{
name: '[instanceNormalization] Test with a non-default epsilon value.',
input: {dataType: 'float32', shape: [1, 2, 3, 4]},
options: {
epsilon: 1e-4,
},
output: {dataType: 'float32', shape: [1, 2, 3, 4]}
},
{
name: '[instanceNormalization] Test with layout=nhwc.',
input: {dataType: 'float32', shape: [1, 2, 3, 4]},
options: {
layout: 'nhwc',
scale: {dataType: 'float32', shape: [4]},
bias: {dataType: 'float32', shape: [4]},
},
output: {dataType: 'float32', shape: [1, 2, 3, 4]}
},
{
name: '[instanceNormalization] Test when the input data type is float16.',
input: {dataType: 'float16', shape: [1, 2, 3, 4]},
output: {dataType: 'float16', shape: [1, 2, 3, 4]},
options: {label}
},
{
name: '[instanceNormalization] Throw if the input is not a 4-D tensor.',
input: {dataType: 'float32', shape: [1, 2, 5, 5, 2]},
options: {label}
},
{
name:
'[instanceNormalization] Throw if the input data type is not one of floating point types.',
input: {dataType: 'int32', shape: [1, 2, 5, 5]},
options: {label}
},
{
name:
'[instanceNormalization] Throw if the scale data type is not the same as the input data type.',
input: {dataType: 'float16', shape: [1, 2, 5, 5]},
options: {
scale: {dataType: 'float32', shape: [2]},
label: label,
},
},
{
name:
'[instanceNormalization] Throw if the scale operand is not a 1-D tensor.',
input: {dataType: 'float32', shape: [1, 2, 5, 5]},
options: {
scale: {dataType: 'float32', shape: [2, 1]},
label: label,
},
},
{
name:
'[instanceNormalization] Throw if the size of scale operand is not equal to the size of the feature dimension of the input with layout=nhwc.',
input: {dataType: 'float32', shape: [1, 2, 5, 5]},
options: {
layout: 'nhwc',
scale: {dataType: 'float32', shape: [2]},
label: label,
},
},
{
name:
'[instanceNormalization] Throw if the size of scale operand is not equal to the size of the feature dimension of the input with layout=nchw.',
input: {dataType: 'float32', shape: [1, 5, 5, 2]},
options: {
layout: 'nchw',
scale: {dataType: 'float32', shape: [2]},
label: label,
},
},
{
name:
'[instanceNormalization] Throw if the bias data type is not the same as the input data type.',
input: {dataType: 'float16', shape: [1, 2, 5, 5]},
options: {
bias: {dataType: 'float32', shape: [2]},
label: label,
},
},
{
name:
'[instanceNormalization] Throw if the bias operand is not a 1-D tensor.',
input: {dataType: 'float32', shape: [1, 2, 5, 5]},
options: {
scale: {dataType: 'float32', shape: [2, 1]},
label: label,
},
},
{
name:
'[instanceNormalization] Throw if the size of bias operand is not equal to the size of the feature dimension of the input with layout=nhwc.',
input: {dataType: 'float32', shape: [1, 2, 5, 5]},
options: {
layout: 'nhwc',
bias: {dataType: 'float32', shape: [2]},
label: label,
},
},
{
name:
'[instanceNormalization] Throw if the size of bias operand is not equal to the size of the feature dimension of the input with layout=nchw.',
input: {dataType: 'float32', shape: [1, 5, 5, 2]},
options: {
layout: 'nchw',
bias: {dataType: 'float32', shape: [2]},
label: label,
},
},
];
tests.forEach(
test => promise_test(async t => {
const builder = new MLGraphBuilder(context);
const input = builder.input('input', test.input);
if (test.options && test.options.bias) {
test.options.bias = builder.input('bias', test.options.bias);
}
if (test.options && test.options.scale) {
test.options.scale = builder.input('scale', test.options.scale);
}
if (test.output &&
context.opSupportLimits()
.instanceNormalization.input.dataTypes.includes(
test.input.dataType)) {
const output = builder.instanceNormalization(input, test.options);
assert_equals(output.dataType, test.output.dataType);
assert_array_equals(output.shape, test.output.shape);
} else {
const regrexp = new RegExp('\\[' + label + '\\]');
assert_throws_with_label(
() => builder.instanceNormalization(input, test.options), regrexp);
}
}, test.name));