48d178cfbb60e5b0004f63b2eca37b2449bb6c67,a2c/a2c.py,Model,__init__,#Model#,26

Before Change

    train_loss = pg_loss + vf_coef * vf_loss

    self.grads_check = grads = tf.gradients(train_loss, params_pi1)
    grads = list(zip(grads, params_pi1))
    trainer = tf.train.RMSPropOptimizer(learning_rate=lr, decay=alpha, epsilon=epsilon)
    _train = trainer.apply_gradients(grads)

After Change

    params = find_trainable_variables("model")
    grads = tf.gradients(loss, params)
    if max_grad_norm is not None:
      grads, _ = tf.clip_by_global_norm(grads, max_grad_norm)
    grads = list(zip(grads, params))
    trainer = tf.train.RMSPropOptimizer(learning_rate=lr, decay=alpha, epsilon=epsilon)
    _train = trainer.apply_gradients(grads)

Instances