webnn/validation_tests/concat.https.any.js

// META: title=validation tests for WebNN API concat operation
// META: global=window
// META: variant=?cpu
// META: variant=?gpu
// META: variant=?npu
// META: script=../resources/utils_validation.js

'use strict';

const label = `concate_123`;
const tests = [
  {
    name: '[concat] Test building Concat with one input.',
    inputs: [{dataType: 'float32', shape: [4, 4, 3]}],
    axis: 2,
    output: {dataType: 'float32', shape: [4, 4, 3]}
  },
  {
    name: '[concat] Test building Concat with two inputs',
    inputs: [
      {dataType: 'float32', shape: [3, 1, 5]},
      {dataType: 'float32', shape: [3, 2, 5]}
    ],
    axis: 1,
    output: {dataType: 'float32', shape: [3, 3, 5]}
  },
  {
    name: '[concat] Test building Concat with three inputs',
    inputs: [
      {dataType: 'float32', shape: [3, 5, 1]},
      {dataType: 'float32', shape: [3, 5, 2]},
      {dataType: 'float32', shape: [3, 5, 3]}
    ],
    axis: 2,
    output: {dataType: 'float32', shape: [3, 5, 6]}
  },
  {
    name: '[concat] Test building Concat with two 1D inputs.',
    inputs:
        [{dataType: 'float32', shape: [1]}, {dataType: 'float32', shape: [1]}],
    axis: 0,
    output: {dataType: 'float32', shape: [2]}
  },
  {
    name: '[concat] Throw if the inputs are empty.',
    axis: 0,
  },
  {
    name: '[concat] Throw if the argument types are inconsistent.',
    inputs: [
      {dataType: 'float32', shape: [1, 1]}, {dataType: 'int32', shape: [1, 1]}
    ],
    axis: 0,
  },
  {
    name: '[concat] Throw if the inputs have different ranks.',
    inputs: [
      {dataType: 'float32', shape: [1, 1]},
      {dataType: 'float32', shape: [1, 1, 1]}
    ],
    axis: 0,
  },
  {
    name:
        '[concat] Throw if the axis is equal to or greater than the size of ranks',
    inputs: [
      {dataType: 'float32', shape: [1, 1]}, {dataType: 'float32', shape: [1, 1]}
    ],
    axis: 2,
  },
  {
    name: '[concat] Throw if concat with two 0-D scalars.',
    inputs:
        [{dataType: 'float32', shape: []}, {dataType: 'float32', shape: []}],
    axis: 0,
  },
  {
    name:
        '[concat] Throw if the inputs have other axes with different sizes except on the axis.',
    inputs: [
      {dataType: 'float32', shape: [1, 1, 1]},
      {dataType: 'float32', shape: [1, 2, 3]}
    ],
    axis: 1,
  },
];

tests.forEach(
    test => promise_test(async t => {
      const builder = new MLGraphBuilder(context);
      let inputs = [];
      if (test.inputs) {
        for (let i = 0; i < test.inputs.length; ++i) {
          inputs[i] = builder.input(`inputs[${i}]`, test.inputs[i]);
        }
      }
      if (test.output) {
        const output = builder.concat(inputs, test.axis);
        assert_equals(output.dataType, test.output.dataType);
        assert_array_equals(output.shape, test.output.shape);
      } else {
        const options = {label};
        const regrexp = new RegExp('\\[' + label + '\\]');
        assert_throws_with_label(
            () => builder.concat(inputs, test.axis, options), regrexp);
      }
    }, test.name));

multi_builder_test(async (t, builder, otherBuilder) => {
  const operandDescriptor = {dataType: 'float32', shape: [2, 2]};

  const inputFromOtherBuilder = otherBuilder.input('input', operandDescriptor);

  const input1 = builder.input('input', operandDescriptor);
  const input2 = builder.input('input', operandDescriptor);
  const input3 = builder.input('input', operandDescriptor);

  assert_throws_js(
      TypeError,
      () => builder.concat([input1, input2, inputFromOtherBuilder, input3]));
}, '[concat] throw if any input is from another builder');

promise_test(async t => {
  const builder = new MLGraphBuilder(context);

  const operandDescriptor = {
    dataType: 'float32',
    shape: [context.opSupportLimits().maxTensorByteLength / 4]
  };
  const input1 = builder.input('input1', operandDescriptor);
  const input2 = builder.input('input2', operandDescriptor);
  const input3 = builder.input('input3', operandDescriptor);

  assert_throws_js(
      TypeError, () => builder.concat(input1, input2, input3));
}, '[concat] throw if the output tensor byte length exceeds limit');