OpenAI基准:尝试在推理过程中使用未初始化的值

时间:2019-05-15 19:18:55

标签: python c++ tensorflow

我正在使用OpenAI基准来构建deepq模型。我通过修改baselines/common/tf_util.py中的代码将模型导出为原始buff格式,如下所示:

def save_state(fname, sess=None):
    from baselines import logger
    logger.warn('save_state method is deprecated, please use save_variables instead')
    sess = sess or get_session()
    dirname = os.path.dirname(fname)
    if any(dirname):
        os.makedirs(dirname, exist_ok=True)

    tf.global_variables_initializer()
    tf.local_variables_initializer()
    saver = tf.train.Saver()
    saver.save(tf.get_default_session(), fname)
    tf.train.write_graph(sess.graph_def, './model', 'deepq_0515.pb', as_text=False)
    tf.train.write_graph(sess.graph_def, './model', 'deepq_0515.pbtxt')

然后我在C ++中加载模型以执行推理。模型很好,很好。但是,当我这样做时:

tensorflow::Status status =
    session_->Run({{"deepq/observation", input_tensor}}, {"deepq/Select"}, {}, &output_tensor);

我遇到以下错误。知道我错过了什么吗?

precondition:Attempting to use uninitialized value deepq/q_func/action_value/fully_connected/biases
     [[{{node deepq/q_func/action_value/fully_connected/biases/read}}]]
 :
precondition: Attempting to use uninitialized value deepq/q_func/action_value/fully_connected_1/weights
     [[{{node deepq/q_func/action_value/fully_connected_1/weights/read}}]]

0 个答案:

没有答案