Argthtjtr

I'm new to tensorflow but trying to run an older script that was uploaded to github. At the moment I'm stuck because tensorflow is trying to use a variable that already exists and I can't figure out where to add reuse. Any thoughts / suggestions would be really helpful! Thanks!

Code:

self.t_vars = tf.trainable_variables()



self.q_vars = [var for var in self.t_vars if (self.model_name+'_q_') in var.name]

self.g_vars = [var for var in self.t_vars if (self.model_name+'_g_') in var.name]

self.d_vars = [var for var in self.t_vars if (self.model_name+'_d_') in var.name]

self.both_vars = self.q_vars+self.g_vars

#self.vae_vars = self.q_vars # in this version, g_vars don't concern vae_loss



# clip gradients

d_opt_real_grads, _ = tf.clip_by_global_norm(tf.gradients(self.d_loss_real, self.d_vars), self.grad_clip)

d_opt_grads, _ = tf.clip_by_global_norm(tf.gradients(self.d_loss, self.d_vars), self.grad_clip)

g_opt_grads, _ = tf.clip_by_global_norm(tf.gradients(self.balanced_loss, self.both_vars), self.grad_clip)

vae_opt_grads, _ = tf.clip_by_global_norm(tf.gradients(self.vae_loss, self.q_vars), self.grad_clip)



d_real_optimizer = tf.train.AdamOptimizer(self.learning_rate_d, beta1=self.beta1)

d_optimizer = tf.train.AdamOptimizer(self.learning_rate_d, beta1=self.beta1)

g_optimizer = tf.train.AdamOptimizer(self.learning_rate_g, beta1=self.beta1)

vae_optimizer = tf.train.AdamOptimizer(self.learning_rate_vae, beta1=self.beta1)



self.d_opt_real = d_real_optimizer.apply_gradients(zip(d_opt_real_grads, self.d_vars))

self.d_opt = d_optimizer.apply_gradients(zip(d_opt_grads, self.d_vars))

self.g_opt = g_optimizer.apply_gradients(zip(g_opt_grads, self.both_vars))

self.vae_opt = vae_optimizer.apply_gradients(zip(vae_opt_grads, self.q_vars))

Error:
ValueError: Variable cppnvae_d_h0_conv/w/Adam/ already exists, disallowed. Did you mean to set reuse=True or reuse=tf.AUTO_REUSE in VarScope?