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

Before Change

    vf_loss = tf.reduce_mean(mse(tf.squeeze(train_model.vf), R))

    self.params = params = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope="model")

    self.params_common = params_common = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope="model/common")

After Change

    loss = pg_loss - entropy*ent_coef + vf_loss * vf_coef

    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)

Instances