Loss of WGAN does not converge

[Curry-whs]

The loss curve of my WGAN does not converge all the time, and I do not know why. Could anyone help me?I have changed the whole model many times, and the best result is that it starts to converge and collapse at 250peochs,the loss curve just like this :

my code looks like this:

So how to solve this problem?

[tf2gan]

Have you solved the problem? My WGAN loss is also not convergent

[senu-et]

Gradient clipping instead of weight clipping

for WGAN model all the gradients should clip to (-0.01,0.01)

    for idx,grad in enumerate(gradients_of_d) :
        gradients_of_d[idx]=tf.clip_by_value(grad,-0.01,0.01)