|
|
@@ -150,7 +150,7 @@ class TensorFlowGraph {
|
|
|
if (tensor) {
|
|
|
this._initializerMap[node.input[0]] = "-";
|
|
|
tensor._id = node.output[0]; // TODO update tensor id
|
|
|
- tensor._title = 'Constant Identity';
|
|
|
+ tensor._title = 'Identity Constant';
|
|
|
this._initializerMap[node.output[0]] = tensor;
|
|
|
}
|
|
|
}
|
|
|
@@ -402,9 +402,15 @@ class TensorFlowAttribute {
|
|
|
}
|
|
|
return list.type.map((type) => TensorFlowTensor.formatDataType(type)).join(', ');
|
|
|
}
|
|
|
+ else if (list.shape && list.shape.length > 0) {
|
|
|
+ if (list.shape.length > 65536) {
|
|
|
+ return "Too large to render.";
|
|
|
+ }
|
|
|
+ return list.shape.map((shape) => TensorFlowTensor.formatTensorShape(shape)).join(', ');
|
|
|
+ }
|
|
|
}
|
|
|
debugger;
|
|
|
- return '?';
|
|
|
+ return '';
|
|
|
}
|
|
|
|
|
|
get hidden() {
|
|
|
@@ -444,7 +450,75 @@ class TensorFlowTensor {
|
|
|
}
|
|
|
|
|
|
get value() {
|
|
|
- return '?';
|
|
|
+ if (!this._tensor.dtype) {
|
|
|
+ return 'Tensor has no data type.';
|
|
|
+ }
|
|
|
+ if (!this._tensor.tensorShape) {
|
|
|
+ return 'Tensor has no dimensions.';
|
|
|
+ }
|
|
|
+
|
|
|
+ switch (this._tensor.dtype) {
|
|
|
+ case tensorflow.DataType.DT_FLOAT:
|
|
|
+ if (this._tensor.tensorContent && this._tensor.tensorContent.length > 0) {
|
|
|
+ this._rawData = new DataView(this._tensor.tensorContent.buffer, this._tensor.tensorContent.byteOffset, this._tensor.tensorContent.byteLength)
|
|
|
+ }
|
|
|
+ else {
|
|
|
+ return 'Tensor data is empty.';
|
|
|
+ }
|
|
|
+ break;
|
|
|
+ default:
|
|
|
+ debugger;
|
|
|
+ return 'Tensor data type is not implemented.';
|
|
|
+ }
|
|
|
+
|
|
|
+ this._index = 0;
|
|
|
+ this._count = 0;
|
|
|
+ var result = this.read(0);
|
|
|
+ delete this._index;
|
|
|
+ delete this._count;
|
|
|
+ delete this._data;
|
|
|
+ delete this._rawData;
|
|
|
+
|
|
|
+ return JSON.stringify(result, null, 4);
|
|
|
+ }
|
|
|
+
|
|
|
+ read(dimension) {
|
|
|
+ var results = [];
|
|
|
+ var dim = this._tensor.tensorShape.dim[dimension];
|
|
|
+ var size = dim.size;
|
|
|
+ if (dimension == this._tensor.tensorShape.dim.length - 1) {
|
|
|
+ for (var i = 0; i < size; i++) {
|
|
|
+ if (this._count > 10000) {
|
|
|
+ results.push('...');
|
|
|
+ return results;
|
|
|
+ }
|
|
|
+ if (this._data) {
|
|
|
+ results.push(this._data[this._index++]);
|
|
|
+ }
|
|
|
+ else {
|
|
|
+ if (this._rawData) {
|
|
|
+ switch (this._tensor.dtype)
|
|
|
+ {
|
|
|
+ case tensorflow.DataType.DT_FLOAT:
|
|
|
+ results.push(this._rawData.getFloat32(this._index, true));
|
|
|
+ this._index += 4;
|
|
|
+ this._count++;
|
|
|
+ break;
|
|
|
+ }
|
|
|
+ }
|
|
|
+ }
|
|
|
+ }
|
|
|
+ }
|
|
|
+ else {
|
|
|
+ for (var j = 0; j < size; j++) {
|
|
|
+ if (this._count > 10000) {
|
|
|
+ results.push('...');
|
|
|
+ return results;
|
|
|
+ }
|
|
|
+ results.push(this.read(dimension + 1));
|
|
|
+ }
|
|
|
+ }
|
|
|
+ return results;
|
|
|
}
|
|
|
|
|
|
static formatTensorType(tensor) {
|
Lutz Roeder