// META: title=validation tests for WebNN API prelu operation
// META: global=window,dedicatedworker
// META: variant=?cpu
// META: variant=?gpu
// META: variant=?npu
// META: script=../resources/utils_validation.js
'use strict';
validateTwoInputsFromMultipleBuilders('prelu');
const tests = [
{
name:
'[prelu] Test slope\'s shape = [3, 2, 5] which is the same as input\'s shape.',
input: {dataType: 'float32', dimensions: [3, 2, 5]},
slope: {dataType: 'float32', dimensions: [3, 2, 5]},
output: {dataType: 'float32', dimensions: [3, 2, 5]},
},
{
name:
'[prelu] Test slope\'s shape = [5] which is unidirectionally broadcastable to input\'s shape.',
input: {dataType: 'float32', dimensions: [3, 2, 5]},
slope: {dataType: 'float32', dimensions: [5]},
output: {dataType: 'float32', dimensions: [3, 2, 5]},
},
{
name:
'[prelu] Test slope\'s shape = [] which is unidirectionally broadcastable to input\'s shape.',
input: {dataType: 'float32', dimensions: [3, 2, 5]},
slope: {dataType: 'float32', dimensions: []},
output: {dataType: 'float32', dimensions: [3, 2, 5]},
},
{
name:
'[prelu] Test slope\'s shape = [2, 5] which is unidirectionally broadcastable to input\'s shape.',
input: {dataType: 'float32', dimensions: [3, 2, 5]},
slope: {dataType: 'float32', dimensions: [2, 5]},
output: {dataType: 'float32', dimensions: [3, 2, 5]},
},
{
name:
'[prelu] Throw if the shape of slope is not broadcastable to the shape of input.',
input: {dataType: 'float32', dimensions: [3, 2, 5]},
slope: {dataType: 'float32', dimensions: [2]},
},
{
name:
'[prelu] Throw if the data type of slope does not match the data type of input.',
input: {dataType: 'float32', dimensions: [3, 2, 5]},
slope: {dataType: 'int32', dimensions: [3, 2, 5]},
},
];
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 slope = builder.input(
'input',
{dataType: test.slope.dataType, dimensions: test.slope.dimensions});
if (test.output) {
const output = builder.prelu(input, slope);
assert_equals(output.dataType(), test.output.dataType);
assert_array_equals(output.shape(), test.output.dimensions);
} else {
const label = 'prelu_123';
const options = {label};
const regrexp = new RegExp('\\[' + label + '\\]');
assert_throws_with_label(
() => builder.prelu(input, slope, options), regrexp);
}
}, test.name));
promise_test(async t => {
for (let dataType of allWebNNOperandDataTypes) {
if (!context.opSupportLimits().input.dataTypes.includes(dataType)) {
continue;
}
const builder = new MLGraphBuilder(context);
const dimensions = [1];
const input = builder.input(`input`, {dataType, dimensions});
if (context.opSupportLimits().prelu.input.dataTypes.includes(dataType)) {
const output = builder.prelu(input, input);
assert_equals(output.dataType(), dataType);
assert_array_equals(output.shape(), dimensions);
} else {
assert_throws_js(TypeError, () => builder.prelu(input, input));
}
}
}, `[prelu] Test prelu with all of the data types.`);