Convert MLOperand methods into readonly attributes

Adapt to spec change at webmachinelearning/webnn#774

face_recognition/facenet_nchw.js

@@ -52,8 +52,8 @@ export class FaceNetNchw {
     if (options.autoPad == 'same-upper') {
       options.padding =
         computePadding2DForAutoPad(
-            /* nchw */[input.shape()[2], input.shape()[3]],
-            /* oihw */[weights.shape()[2], weights.shape()[3]],
+            /* nchw */[input.shape[2], input.shape[3]],
+            /* oihw */[weights.shape[2], weights.shape[3]],
             options.strides, options.dilations, options.autoPad);
     }
     const conv2d = this.builder_.conv2d(input, weights, options);
@@ -266,7 +266,7 @@ export class FaceNetNchw {
 
     const averagePool = this.builder_.averagePool2d(await block8_6);
     // Use reshape to implement squeeze(averagePool, {axes: [2, 3]});
-    const squeezed_shape = averagePool.shape();
+    const squeezed_shape = averagePool.shape;
     squeezed_shape.splice(2, 2);
     const squeeze = this.builder_.reshape(averagePool, squeezed_shape);
     const gemm = await this.buildGemm_(squeeze);

face_recognition/facenet_nhwc.js

@@ -54,8 +54,8 @@ export class FaceNetNhwc {
     if (options.autoPad == 'same-upper') {
       options.padding =
         computePadding2DForAutoPad(
-            /* nwhc */[input.shape()[1], input.shape()[2]],
-            /* ohwi */[weights.shape()[1], weights.shape()[2]],
+            /* nwhc */[input.shape[1], input.shape[2]],
+            /* ohwi */[weights.shape[1], weights.shape[2]],
             options.strides, options.dilations, options.autoPad);
     }
     const conv2d = this.builder_.conv2d(input, weights, options);

facial_landmark_detection/ssd_mobilenetv2_face_nchw.js

@@ -69,8 +69,8 @@ ${nameArray[1]}`;
     const weights = buildConstantByNpy(this.builder_, weightsName);
     const biasName = prefix + biasSuffix;
     const bias = buildConstantByNpy(this.builder_, biasName);
-    const inputShape = (await input).shape();
-    const weightsShape = (await weights).shape();
+    const inputShape = (await input).shape;
+    const weightsShape = (await weights).shape;
     options.padding = computePadding2DForAutoPad(
         /* nchw */[inputShape[2], inputShape[3]],
         /* oihw */[weightsShape[2], weightsShape[3]],

facial_landmark_detection/ssd_mobilenetv2_face_nhwc.js

@@ -82,8 +82,8 @@ ${nameArray[1]}`;
       options.filterLayout = 'ihwo';
     }
     options.bias = await bias;
-    const inputShape = (await input).shape();
-    const weightsShape = (await weights).shape();
+    const inputShape = (await input).shape;
+    const weightsShape = (await weights).shape;
     options.padding = computePadding2DForAutoPad(
         /* nhwc */[inputShape[1], inputShape[2]],
         /* ohwi or ihwo */[weightsShape[1], weightsShape[2]],

image_classification/mobilenet_nhwc.js

@@ -36,8 +36,8 @@ export class MobileNetV2Nhwc {
     if (options.autoPad == 'same-upper') {
       options.padding =
         computePadding2DForAutoPad(
-            /* nwhc */[await input.shape()[1], await input.shape()[2]],
-            /* ohwi or ihwo */[weights.shape()[1], weights.shape()[2]],
+            /* nwhc */[await input.shape[1], await input.shape[2]],
+            /* ohwi or ihwo */[weights.shape[1], weights.shape[2]],
             options.strides, options.dilations, options.autoPad);
     }
     const conv2d = this.builder_.conv2d(await input, weights, options);

image_classification/resnet50v2_nhwc.js

@@ -54,8 +54,8 @@ export class ResNet50V2Nhwc {
     if (options.autoPad == 'same-upper') {
       options.padding =
         computePadding2DForAutoPad(
-            /* nwhc */[await input.shape()[1], await input.shape()[2]],
-            /* ohwi */[weights.shape()[1], weights.shape()[2]],
+            /* nwhc */[await input.shape[1], await input.shape[2]],
+            /* ohwi */[weights.shape[1], weights.shape[2]],
             options.strides, options.dilations, options.autoPad);
     }
     const conv2d = this.builder_.conv2d(await input, weights, options);
@@ -144,7 +144,7 @@ export class ResNet50V2Nhwc {
     const pool = this.builder_.maxPool2d(
         conv1, {windowDimensions, strides, layout,
           padding: computePadding2DForAutoPad(
-              /* nhwc */ [conv1.shape()[1], conv1.shape()[2]],
+              /* nhwc */ [conv1.shape[1], conv1.shape[2]],
               windowDimensions, strides, /* dilations */ undefined,
               'same-upper')});
     // Block 1

image_classification/squeezenet_nhwc.js

@@ -35,8 +35,8 @@ export class SqueezeNetNhwc {
     if (options.autoPad == 'same-upper') {
       options.padding =
         computePadding2DForAutoPad(
-            /* nwhc */[await input.shape()[1], await input.shape()[2]],
-            /* ohwi */[weights.shape()[1], weights.shape()[2]],
+            /* nwhc */[await input.shape[1], await input.shape[2]],
+            /* ohwi */[weights.shape[1], weights.shape[2]],
             options.strides, options.dilations, options.autoPad);
     }
     const conv2d = this.builder_.conv2d(await input, weights, options);

nnotepad/js/nnotepad.js

@@ -37,16 +37,16 @@ const kArgTypeOperand = 3;
 
 class WebNNUtil {
   static bufferForOperand(operand) {
-    const size = [...operand.shape()].reduce((a, b) => a * b, 1);
-    const ctor = WebNNUtil.dataTypeToBufferType(operand.dataType());
+    const size = [...operand.shape].reduce((a, b) => a * b, 1);
+    const ctor = WebNNUtil.dataTypeToBufferType(operand.dataType);
     return Reflect.construct(ctor, [size]);
   }
 
   static async tensorForOperand(operand, context) {
     const desc = {
-      dataType: operand.dataType(),
-      dimensions: operand.shape(),
-      shape: operand.shape(),
+      dataType: operand.dataType,
+      dimensions: operand.shape,
+      shape: operand.shape,
       usage: MLTensorUsage.READ,
       readable: true,
     };
@@ -613,9 +613,9 @@ export class NNotepad {
 
     return outputOperands.map(
         (op, index) => ({
-          dataType: op.dataType(),
-          dimensions: op.shape(),
-          shape: op.shape(),
+          dataType: op.dataType,
+          dimensions: op.shape,
+          shape: op.shape,
           buffer: maybeProxyForFloat16Array(outputBuffers[`output-${index}`]),
         }));
   }

nsnet2/nsnet2.js

@@ -58,7 +58,7 @@ export class NSNet2 {
     const [gru94, gru93] = this.builder_.gru(transpose31, weight192, recurrentWeight193, frames, this.hiddenSize,
         {bias: bias194, recurrentBias: recurrentBias194, initialHiddenState: initialState92, returnSequence: true});
     // Use reshape to implement squeeze(gru93, {axes: [1]});
-    const squeeze95Shape = gru93.shape();
+    const squeeze95Shape = gru93.shape;
     squeeze95Shape.splice(1, 1);
     const squeeze95 = this.builder_.reshape(gru93, squeeze95Shape);
     const initialState155 = this.builder_.input('initialState155', initialStateDesc);
@@ -89,7 +89,7 @@ export class NSNet2 {
     const [gru157, gru156] = this.builder_.gru(squeeze95, weight212, recurrentWeight213, frames, this.hiddenSize,
         {bias: bias214, recurrentBias: recurrentBias214, initialHiddenState: initialState155, returnSequence: true});
     // Use reshape to implement squeeze(gru156, {axes: [1]});
-    const squeeze158Shape = gru156.shape();
+    const squeeze158Shape = gru156.shape;
     squeeze158Shape.splice(1, 1);
     const squeeze158 = this.builder_.reshape(gru156, squeeze158Shape);
     const transpose159 = this.builder_.transpose(squeeze158, {permutation: [1, 0, 2]});

object_detection/ssd_mobilenetv1_nchw.js

@@ -69,8 +69,8 @@ ${nameArray[1]}_BatchNorm_batchnorm`;
     const bias = await buildConstantByNpy(
         this.builder_, biasName, this.targetDataType_);
     options.padding = computePadding2DForAutoPad(
-        /* nchw */[input.shape()[2], input.shape()[3]],
-        /* oihw */[weights.shape()[2], weights.shape()[3]],
+        /* nchw */[input.shape[2], input.shape[3]],
+        /* oihw */[weights.shape[2], weights.shape[3]],
         options.strides, options.dilations, 'same-upper');
     options.bias = bias;
     const conv2d = this.builder_.conv2d(input, weights, options);

object_detection/ssd_mobilenetv1_nhwc.js

@@ -76,8 +76,8 @@ ${nameArray[1]}_BatchNorm_batchnorm`;
     }
     options.bias = bias;
     options.padding = computePadding2DForAutoPad(
-        /* nhwc */[input.shape()[1], input.shape()[2]],
-        /* ohwi or ihwo */[weights.shape()[1], weights.shape()[2]],
+        /* nhwc */[input.shape[1], input.shape[2]],
+        /* ohwi or ihwo */[weights.shape[1], weights.shape[2]],
         options.strides, options.dilations, 'same-upper');
     const conv2d = this.builder_.conv2d(input, weights, options);
     if (relu6) {

object_detection/tiny_yolov2_nchw.js

@@ -38,8 +38,8 @@ export class TinyYoloV2Nchw {
         this.builder_, weightName, this.targetDataType_);
     const options = {autoPad: 'same-upper'};
     options.padding = computePadding2DForAutoPad(
-        /* nchw */[input.shape()[2], input.shape()[3]],
-        /* oihw */[weight.shape()[2], weight.shape()[3]],
+        /* nchw */[input.shape[2], input.shape[3]],
+        /* oihw */[weight.shape[2], weight.shape[3]],
         options.strides, options.dilations, 'same-upper');
     options.bias = await buildConstantByNpy(
         this.builder_, biasName, this.targetDataType_);
@@ -53,7 +53,7 @@ export class TinyYoloV2Nchw {
 
   buildMaxPool2d_(input, options) {
     options.padding = computePadding2DForAutoPad(
-        /* nchw */[input.shape()[2], input.shape()[3]],
+        /* nchw */[input.shape[2], input.shape[3]],
         options.windowDimensions,
         options.strides, options.dilations, 'same-upper');
     return this.builder_.maxPool2d(input, options);

object_detection/tiny_yolov2_nhwc.js

@@ -35,8 +35,8 @@ export class TinyYoloV2Nhwc {
     };
     options.bias = bias;
     options.padding = computePadding2DForAutoPad(
-        /* nhwc */[input.shape()[1], input.shape()[2]],
-        /* ohwi */[weights.shape()[1], weights.shape()[2]],
+        /* nhwc */[input.shape[1], input.shape[2]],
+        /* ohwi */[weights.shape[1], weights.shape[2]],
         options.strides, options.dilations, 'same-upper');
     let conv = this.builder_.conv2d(input, weights, options);
     if (leakyRelu) {
@@ -48,7 +48,7 @@ export class TinyYoloV2Nhwc {
 
   buildMaxPool2d_(input, options) {
     options.padding = computePadding2DForAutoPad(
-        /* nhwc */[input.shape()[1], input.shape()[2]],
+        /* nhwc */[input.shape[1], input.shape[2]],
         options.windowDimensions,
         options.strides, options.dilations, 'same-upper');
     return this.builder_.maxPool2d(input, options);

semantic_segmentation/deeplabv3_mnv2_nhwc.js

@@ -45,8 +45,8 @@ export class DeepLabV3MNV2Nhwc {
       options.filterLayout = 'ohwi';
     }
     options.padding = computePadding2DForAutoPad(
-        /* nhwc */[input.shape()[1], input.shape()[2]],
-        /* ohwi or ihwo */[weights.shape()[1], weights.shape()[2]],
+        /* nhwc */[input.shape[1], input.shape[2]],
+        /* ohwi or ihwo */[weights.shape[1], weights.shape[2]],
         options.strides, options.dilations, 'same-upper');
     options.bias = bias;
     const conv2d = this.builder_.conv2d(input, weights, options);

style_transfer/fast_style_transfer_net.js

@@ -26,8 +26,8 @@ export class FastStyleTransferNet {
   buildInstanceNormalization_(conv2D, variableMul, variableAdd) {
     if ('instanceNormalization' in this.builder_) {
       // Use reshape to implement squeeze(variableMul); and squeeze(variableAdd);
-      const mulShape = variableMul.shape().filter((dim) => dim !==1);
-      const addShape = variableAdd.shape().filter((dim) => dim !==1);
+      const mulShape = variableMul.shape.filter((dim) => dim !==1);
+      const addShape = variableAdd.shape.filter((dim) => dim !==1);
       const mulSqueeze = this.builder_.reshape(variableMul, mulShape);
       const addSqueeze = this.builder_.reshape(variableAdd, addShape);
       return this.builder_.instanceNormalization(conv2D, {scale: mulSqueeze, bias: addSqueeze});