18
18
/**
19
19
* This file loads a pre-trained generator part of an ACGAN and demonstrates
20
20
* the generation of fake MNIST images.
21
- *
21
+ *
22
22
* The pre-trained generator model may come from either of the two sources:
23
23
* 1. Running the traning script `gan.js` in the same folder.
24
24
* 2. A hosted model, via HTTPS requests.
@@ -105,14 +105,14 @@ async function generateAndVisualizeImages(generator) {
105
105
return tf . concat ( tf . unstack ( generatedImages ) , 1 ) ;
106
106
} ) ;
107
107
108
- await tf . toPixels ( combinedFakes , fakeCanvas ) ;
108
+ await tf . browser . toPixels ( combinedFakes , fakeCanvas ) ;
109
109
tf . dispose ( combinedFakes ) ;
110
110
}
111
111
112
112
/** Refresh examples of real MNIST images. */
113
113
async function drawReals ( ) {
114
114
const combinedReals = sampleFromMnistData ( 10 ) ;
115
- await tf . toPixels ( combinedReals , realCanvas ) ;
115
+ await tf . browser . toPixels ( combinedReals , realCanvas ) ;
116
116
tf . dispose ( combinedReals ) ;
117
117
}
118
118
@@ -126,7 +126,7 @@ let latentSliders;
126
126
*/
127
127
function createSliders ( generator ) {
128
128
const latentDims = generator . inputs [ 0 ] . shape [ 1 ] ;
129
- latentSliders = [ ] ;
129
+ latentSliders = [ ] ;
130
130
for ( let i = 0 ; i < latentDims ; ++ i ) {
131
131
const slider = document . createElement ( 'input' ) ;
132
132
slider . setAttribute ( 'type' , 'range' ) ;
0 commit comments