// META: title=validation tests for WebNN API batchNormalization operation
// META: global=window,dedicatedworker
// META: variant=?cpu
// META: variant=?gpu
// META: variant=?npu
// META: script=../resources/utils_validation.js
'use strict';
let meanIndex = 0;
let varianceIndex = 0;
const kExampleInputDescriptor = {
dataType: 'float32',
dimensions: [2, 2]
};
// 1D tensor descriptor which may be used for `mean`, `variance`, `scale`, or
// `bias` inputs.
const kExample1DTensorDescriptor = {
dataType: 'float32',
dimensions: [kExampleInputDescriptor.dimensions[/* axis */ 1]]
};
multi_builder_test(async (t, builder, otherBuilder) => {
const inputFromOtherBuilder =
otherBuilder.input('input', kExampleInputDescriptor);
const mean = builder.input('mean', kExample1DTensorDescriptor);
const variance = builder.input('variance', kExample1DTensorDescriptor);
assert_throws_js(
TypeError, () => builder.batchNormalization(inputFromOtherBuilder, mean, variance));
}, '[batchNormalization] throw if input is from another builder');
multi_builder_test(async (t, builder, otherBuilder) => {
const meanFromOtherBuilder =
otherBuilder.input('mean', kExample1DTensorDescriptor);
const input = builder.input('input', kExampleInputDescriptor);
const variance = builder.input('variance', kExample1DTensorDescriptor);
assert_throws_js(
TypeError,
() => builder.batchNormalization(input, meanFromOtherBuilder, variance));
}, '[batchNormalization] throw if mean is from another builder');
multi_builder_test(async (t, builder, otherBuilder) => {
const varianceFromOtherBuilder =
otherBuilder.input('variance', kExample1DTensorDescriptor);
const input = builder.input('input', kExampleInputDescriptor);
const mean = builder.input('mean', kExample1DTensorDescriptor);
assert_throws_js(
TypeError,
() => builder.batchNormalization(input, mean, varianceFromOtherBuilder));
}, '[batchNormalization] throw if variance 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);
const mean = builder.input('mean', kExample1DTensorDescriptor);
const variance = builder.input('variance', kExample1DTensorDescriptor);
assert_throws_js(
TypeError,
() => builder.batchNormalization(input, mean, variance, options));
}, '[batchNormalization] throw if scale option is from another builder');
multi_builder_test(async (t, builder, otherBuilder) => {
const biasFromOtherBuilder =
otherBuilder.input('bias', kExample1DTensorDescriptor);
const options = {scale: biasFromOtherBuilder};
const input = builder.input('input', kExampleInputDescriptor);
const mean = builder.input('mean', kExample1DTensorDescriptor);
const variance = builder.input('variance', kExample1DTensorDescriptor);
assert_throws_js(
TypeError,
() => builder.batchNormalization(input, mean, variance, options));
}, '[batchNormalization] throw if bias option is from another builder');
const label = `batchNormalization_?_123`;
const tests = [
{
name: '[batchNormalization] Test with default options.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [2]},
variance: {dataType: 'float32', dimensions: [2]},
output: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
},
{
name: '[batchNormalization] Test with axis = 2 and epsilon = 0.0001.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [5]},
variance: {dataType: 'float32', dimensions: [5]},
options: {
axis: 2,
epsilon: 1e-4, // 1e-5 is the default value of epsilon.
},
output: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
},
{
name:
'[batchNormalization] Throw if the input data type is not one of floating point types.',
input: {dataType: 'int32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'int32', dimensions: [2]},
variance: {dataType: 'int32', dimensions: [2]},
options: {
label: label,
},
},
{
name:
'[batchNormalization] Throw if the mean data type is not the same as the input data type.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float16', dimensions: [2]},
variance: {dataType: 'float32', dimensions: [2]},
options: {
label: label,
},
},
{
name: '[batchNormalization] Throw if the mean operand is not a 1-D tensor.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [1, 2]},
variance: {dataType: 'float32', dimensions: [2]},
options: {
label: label,
},
},
{
name:
'[batchNormalization] Throw if the size of mean operand is not equal to the size of the input dimension denoted by axis.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [3]},
variance: {dataType: 'float32', dimensions: [2]},
options: {
axis: 1,
label: label,
},
},
{
name:
'[batchNormalization] Throw if the variance data type is not the same as the input data type.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [2]},
variance: {dataType: 'float16', dimensions: [2]},
options: {
label: label,
},
},
{
name:
'[batchNormalization] Throw if the variance operand is not a 1-D tensor.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [2]},
variance: {dataType: 'float32', dimensions: [2, 2]},
options: {
label: label,
},
},
{
name:
'[batchNormalization] Throw if the size of variance operand is not equal to the size of the input dimension denoted by axis.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [5]},
variance: {dataType: 'float32', dimensions: [2]},
options: {
axis: 2,
label: label,
},
},
{
name:
'[batchNormalization] Throw if the scale data type is not the same as the input data type.',
input: {dataType: 'float16', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float16', dimensions: [2]},
variance: {dataType: 'float16', dimensions: [2]},
options: {
scale: {dataType: 'float32', dimensions: [2]},
label: label,
},
},
{
name:
'[batchNormalization] Throw if the scale operand is not a 1-D tensor.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [2]},
variance: {dataType: 'float32', dimensions: [2]},
options: {
scale: {dataType: 'float32', dimensions: [2, 1]},
label: label,
},
},
{
name:
'[batchNormalization] Throw if the size of scale operand is not equal to the size of the input dimension denoted by axis.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [5]},
variance: {dataType: 'float32', dimensions: [5]},
options: {
axis: 2,
scale: {dataType: 'float32', dimensions: [2]},
label: label,
},
},
{
name:
'[batchNormalization] Throw if the bias data type is not the same as the input data type.',
input: {dataType: 'float16', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float16', dimensions: [2]},
variance: {dataType: 'float16', dimensions: [2]},
options: {
bias: {dataType: 'float32', dimensions: [2]},
label: label,
},
},
{
name: '[batchNormalization] Throw if the bias operand is not a 1-D tensor.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [2]},
variance: {dataType: 'float32', dimensions: [2]},
options: {
bias: {dataType: 'float32', dimensions: [2, 1]},
label: label,
},
},
{
name:
'[batchNormalization] Throw if the size of bias operand is not equal to the size of the input dimension denoted by axis.',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [5]},
variance: {dataType: 'float32', dimensions: [5]},
options: {
axis: 2,
bias: {dataType: 'float32', dimensions: [2]},
label: label,
},
},
{
name:
'[batchNormalization] Throw if the value of axis is not in the range of [0,N-1].',
input: {dataType: 'float32', dimensions: [1, 2, 5, 5]},
mean: {dataType: 'float32', dimensions: [5]},
variance: {dataType: 'float32', dimensions: [5]},
options: {
axis: 4,
label: label,
},
},
];
tests.forEach(
test => promise_test(async t => {
const builder = new MLGraphBuilder(context);
const input = builder.input(
'input',
{dataType: test.input.dataType, dimensions: test.input.dimensions});
const mean = builder.input(
'mean',
{dataType: test.mean.dataType, dimensions: test.mean.dimensions});
const variance = builder.input('variance', {
dataType: test.variance.dataType,
dimensions: test.variance.dimensions
});
if (test.options && test.options.bias) {
test.options.bias = builder.input('bias', {
dataType: test.options.bias.dataType,
dimensions: test.options.bias.dimensions
});
}
if (test.options && test.options.scale) {
test.options.scale = builder.input('scale', {
dataType: test.options.scale.dataType,
dimensions: test.options.scale.dimensions
});
}
if (test.output) {
const output =
builder.batchNormalization(input, mean, variance, test.options);
assert_equals(output.dataType(), test.output.dataType);
assert_array_equals(output.shape(), test.output.dimensions);
} else {
const regrexp = /\[batchNormalization_\?_123\]/;
assert_throws_with_label(
() =>
builder.batchNormalization(input, mean, variance, test.options),
regrexp);
}
}, test.name));
Codebase Browser
chromium