draw icon indicating copy to clipboard operation
draw copied to clipboard
verified publisher icon ericjang

The noise sampling and the Lz loss curve issue

Open johnzlj opened this issue 5 years ago • 0 comments

Ln-44: e=tf.random_normal((batch_size,z_size), mean=0, stddev=1) # Qsampler noise I think it should be placed into the funcion sampleQ. Or else the inference will fail.

However, when I made such a modification, the Lz loss-curve will increase instead of decline, as shown here: image

Why?

johnzlj avatar Mar 28 '20 05:03 johnzlj