Add randomStandardNormal op (#6533)

Fixes #4156

Co-authored-by: Matthew Soulanille <msoulanille@google.com>

tfjs-converter/docs/supported_ops.md

@@ -146,6 +146,7 @@
 |OneHot|oneHot|
 |Ones|ones|
 |OnesLike|onesLike|
+|RandomStandardNormal|RandomStandardNormal|
 |RandomUniform|RandomUniform|
 |Range|range|
 |TruncatedNormal|truncatedNormal|

tfjs-converter/python/tensorflowjs/op_list/creation.json

@@ -129,6 +129,43 @@
       }
     ]
   },
+  {
+    "tfOpName": "RandomStandardNormal",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "T",
+        "name": "T",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
   {
     "tfOpName": "RandomUniform",
     "category": "creation",

tfjs-converter/src/operations/executors/creation_executor.ts

@@ -75,6 +75,13 @@ export const executeOp: InternalOpExecutor =
           return [tfOps.onesLike(
               getParamValue('x', node, tensorMap, context) as Tensor)];
         }
+        case 'RandomStandardNormal': {
+          return [tfOps.randomStandardNormal(
+              getParamValue('shape', node, tensorMap, context) as number[],
+              getParamValue('dtype', node, tensorMap, context) as 'float32' |
+                  'int32',
+              getParamValue('seed', node, tensorMap, context) as number)];
+        }
         case 'RandomUniform': {
           return [tfOps.randomUniform(
               // tslint:disable-next-line:no-any

tfjs-converter/src/operations/executors/creation_executor_test.ts

@@ -173,6 +173,30 @@ describe('creation', () => {
         expect(validateParam(node, creation.json)).toBeTruthy();
       });
     });
+    describe('RandomStandardNormal', () => {
+      it('should call tfOps.randomStandardNormal', () => {
+        spyOn(tfOps, 'randomStandardNormal');
+        node.op = 'RandomStandardNormal';
+        node.inputParams['shape'] = createNumericArrayAttrFromIndex(0);
+        node.inputNames = ['input1'];
+        node.attrParams['dtype'] = createDtypeAttr('float32');
+        node.attrParams['seed'] = createNumberAttr(0);
+
+        executeOp(node, {input1}, context);
+
+        expect(tfOps.randomStandardNormal)
+            .toHaveBeenCalledWith([1, 2, 3], 'float32', 0);
+      });
+      it('should match json def', () => {
+        node.op = 'RandomStandardNormal';
+        node.inputParams['shape'] = createNumericArrayAttrFromIndex(0);
+        node.inputNames = ['input1'];
+        node.attrParams['dtype'] = createDtypeAttr('float32');
+        node.attrParams['seed'] = createNumberAttr(0);
+
+        expect(validateParam(node, creation.json)).toBeTruthy();
+      });
+    });
     describe('RandomUniform', () => {
       it('should call tfOps.randomUniform', () => {
         spyOn(tfOps, 'randomUniform');

tfjs-core/src/ops/ops.ts

@@ -138,6 +138,7 @@ export {prod} from './prod';
 export {rand} from './rand';
 export {randomGamma} from './random_gamma';
 export {randomNormal} from './random_normal';
+export {randomStandardNormal} from './random_standard_normal';
 export {randomUniform} from './random_uniform';
 export {range} from './range';
 export {real} from './real';

tfjs-core/src/ops/random_standard_normal.ts

@@ -0,0 +1,47 @@
+/**
+ * @license
+ * Copyright 2022 Google LLC. All Rights Reserved.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * =============================================================================
+ */
+
+import {Tensor} from '../tensor';
+import {DataType, Rank, ShapeMap} from '../types';
+
+import {op} from './operation';
+import {randomNormal} from './random_normal';
+
+/**
+ * Creates a `tf.Tensor` with values sampled from a normal distribution.
+ *
+ * The generated values will have mean 0 and standard deviation 1.
+ *
+ * ```js
+ * tf.randomStandardNormal([2, 2]).print();
+ * ```
+ *
+ * @param shape An array of integers defining the output tensor shape.
+ * @param dtype The data type of the output.
+ * @param seed The seed for the random number generator.
+ *
+ * @doc {heading: 'Tensors', subheading: 'Random'}
+ */
+function randomStandardNormal_<R extends Rank>(
+    shape: ShapeMap[R], dtype?: 'float32'|'int32', seed?: number): Tensor<R> {
+  if (dtype != null && (dtype as DataType) === 'bool') {
+    throw new Error(`Unsupported data type ${dtype}`);
+  }
+  return randomNormal(shape, 0, 1, dtype, seed);
+}
+
+export const randomStandardNormal = op({randomStandardNormal_});

tfjs-core/src/ops/random_standard_normal_test.ts

@@ -0,0 +1,146 @@
+
+/**
+ * @license
+ * Copyright 2022 Google LLC. All Rights Reserved.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * =============================================================================
+ */
+
+import * as tf from '../index';
+import {ALL_ENVS, describeWithFlags} from '../jasmine_util';
+
+import {expectArrayInMeanStdRange, jarqueBeraNormalityTest} from './rand_util';
+
+describeWithFlags('randomStandardNormal', ALL_ENVS, () => {
+  const SEED = 42;
+  const EPSILON = 0.05;
+
+  it('should return a float32 1D of random standard normal values',
+     async () => {
+       const SAMPLES = 10000;
+
+       // Ensure defaults to float32.
+       let result = tf.randomStandardNormal([SAMPLES], null, SEED);
+       expect(result.dtype).toBe('float32');
+       expect(result.shape).toEqual([SAMPLES]);
+       jarqueBeraNormalityTest(await result.data());
+       expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+
+       result = tf.randomStandardNormal([SAMPLES], 'float32', SEED);
+       expect(result.dtype).toBe('float32');
+       expect(result.shape).toEqual([SAMPLES]);
+       jarqueBeraNormalityTest(await result.data());
+       expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+     });
+
+  it('should return a int32 1D of random standard normal values', async () => {
+    const SAMPLES = 10000;
+    const result = tf.randomStandardNormal([SAMPLES], 'int32', SEED);
+    expect(result.dtype).toBe('int32');
+    expect(result.shape).toEqual([SAMPLES]);
+    jarqueBeraNormalityTest(await result.data());
+    expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+  });
+
+  it('should return a float32 2D of random standard normal values',
+     async () => {
+       const SAMPLES = 100;
+
+       // Ensure defaults to float32.
+       let result = tf.randomStandardNormal([SAMPLES, SAMPLES], null, SEED);
+       expect(result.dtype).toBe('float32');
+       expect(result.shape).toEqual([SAMPLES, SAMPLES]);
+       jarqueBeraNormalityTest(await result.data());
+       expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+
+       result = tf.randomStandardNormal([SAMPLES, SAMPLES], 'float32', SEED);
+       expect(result.dtype).toBe('float32');
+       expect(result.shape).toEqual([SAMPLES, SAMPLES]);
+       jarqueBeraNormalityTest(await result.data());
+       expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+     });
+
+  it('should return a int32 2D of random standard normal values', async () => {
+    const SAMPLES = 100;
+    const result = tf.randomStandardNormal([SAMPLES, SAMPLES], 'int32', SEED);
+    expect(result.dtype).toBe('int32');
+    expect(result.shape).toEqual([SAMPLES, SAMPLES]);
+    jarqueBeraNormalityTest(await result.data());
+    expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+  });
+
+  it('should return a float32 3D of random standard normal values',
+     async () => {
+       const SAMPLES_SHAPE = [20, 20, 20];
+
+       // Ensure defaults to float32.
+       let result = tf.randomStandardNormal(SAMPLES_SHAPE, null, SEED);
+       expect(result.dtype).toBe('float32');
+       expect(result.shape).toEqual(SAMPLES_SHAPE);
+       jarqueBeraNormalityTest(await result.data());
+       expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+
+       result = tf.randomStandardNormal(SAMPLES_SHAPE, 'float32', SEED);
+       expect(result.dtype).toBe('float32');
+       expect(result.shape).toEqual(SAMPLES_SHAPE);
+       jarqueBeraNormalityTest(await result.data());
+       expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+     });
+
+  it('should return a int32 3D of random standard normal values', async () => {
+    const SAMPLES_SHAPE = [20, 20, 20];
+    const result = tf.randomStandardNormal(SAMPLES_SHAPE, 'int32', SEED);
+    expect(result.dtype).toBe('int32');
+    expect(result.shape).toEqual(SAMPLES_SHAPE);
+    jarqueBeraNormalityTest(await result.data());
+    expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+  });
+
+  it('should return a float32 4D of random standard normal values',
+     async () => {
+       const SAMPLES_SHAPE = [10, 10, 10, 10];
+
+       // Ensure defaults to float32.
+       let result = tf.randomStandardNormal(SAMPLES_SHAPE, null, SEED);
+       expect(result.dtype).toBe('float32');
+       expect(result.shape).toEqual(SAMPLES_SHAPE);
+       jarqueBeraNormalityTest(await result.data());
+       expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+
+       result = tf.randomStandardNormal(SAMPLES_SHAPE, 'float32', SEED);
+       expect(result.dtype).toBe('float32');
+       expect(result.shape).toEqual(SAMPLES_SHAPE);
+       jarqueBeraNormalityTest(await result.data());
+       expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+     });
+
+  it('should return a int32 4D of random standard normal values', async () => {
+    const SAMPLES_SHAPE = [10, 10, 10, 10];
+
+    const result = tf.randomStandardNormal(SAMPLES_SHAPE, 'int32', SEED);
+    expect(result.dtype).toBe('int32');
+    expect(result.shape).toEqual(SAMPLES_SHAPE);
+    jarqueBeraNormalityTest(await result.data());
+    expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+  });
+
+  it('should return a int32 5D of random standard normal values', async () => {
+    const SAMPLES_SHAPE = [10, 10, 10, 10, 10];
+
+    const result = tf.randomStandardNormal(SAMPLES_SHAPE, 'int32', SEED);
+    expect(result.dtype).toBe('int32');
+    expect(result.shape).toEqual(SAMPLES_SHAPE);
+    jarqueBeraNormalityTest(await result.data());
+    expectArrayInMeanStdRange(await result.data(), 0, 1, EPSILON);
+  });
+});