Does this code allow generators to act as a hybrid model, capturing one component per generator?

tomwesolowski / madgan

MAD-GAN implementation in TensorFlow

2 stars 1 forks source link

Does this code allow generators to act as a hybrid model, capturing one component per generator? #1

Open Tinamilee opened 5 years ago

Tinamilee commented 5 years ago

def generate(sess, model, params): print("Generating images...") images, labels, probs = None, None, None threshold_probs = np.zeros(params.labels_size) for i in trange(params.nb_generated // params.batch_size): gen_outputs, dis_outputs, gen_labels = sess.run( [model.gen_output, model.dis_fake_output_prob, model.zy], feed_dict={model.training: False})

    Running error  
    AttributeError: 'SmartDict' object has no attribute 'gen_output'
    mode has no attribute 'gen_output',  May I ask why?

tomwesolowski commented 5 years ago

Thanks for reaching out.

What do you mean by 'component'? Regarding the error: It seems this is a bug (probably due to the fact this is not the most recent version of the code I used) I didn't run the code, but the line: https://github.com/tomwesolowski/madgan/blob/master/gan.py#L146 tells me it should be model.gen_output_X where X is generator number.