34ee92135551016c240139f7e078c09a583063fb,softlearning/algorithms/sac.py,SAC,_init_critic_update,#SAC#,205

Before Change

                name="{}_{}_optimizer".format(Q._name, i)
            ) for i, Q in enumerate(self._Qs))
        Q_training_ops = tuple(
            tf.contrib.layers.optimize_loss(
                Q_loss,
                self.global_step,
                learning_rate=self._Q_lr,
                optimizer=Q_optimizer,
                variables=Q.trainable_variables,
                increment_global_step=False,
                summaries=((
                    "loss", "gradients", "gradient_norm", "global_gradient_norm"
                ) if self._tf_summaries else ()))
            for i, (Q, Q_loss, Q_optimizer)
            in enumerate(zip(self._Qs, Q_losses, self._Q_optimizers)))

        self._training_ops.update({"Q": tf.group(Q_training_ops)})

After Change

            ) for i, Q in enumerate(self._Qs))

        Q_training_ops = tuple(
            Q_optimizer.minimize(loss=Q_loss, var_list=Q.trainable_variables)
            for i, (Q, Q_loss, Q_optimizer)
            in enumerate(zip(self._Qs, Q_losses, self._Q_optimizers)))

        self._training_ops.update({"Q": tf.group(Q_training_ops)})

Instances