Question

我发现当我尝试运行一个我在gpu上训练的模型时，现在只在一个tensorflow cpu上使用docker图像，我得到以下error

以下是我保存模型的代码：

def fit(self, save=True):
    saver = tf.train.Saver()
    with tf.Session() as sess:
        if self.restore_path:
            restore_model(self.restore_path, sess, saver)
        else :
            start = tf.global_variables_initializer()
            sess.run(start)

        for epoch in xrange(self.config.max_epochs):
            train_pp = self._run_epoch(self._model, self._data, sess, self._model.trainOp, self.verbose)
            print "Training preplexity for batch {} - {}".format(epoch, train_pp)
            if self._validation_set:
                validation_pp = self._run_epoch(self._model, self._validation_set, sess, verbose=self.verbose)
            print "Validation preplexity for batch  {} - {}".format(epoch, validation_pp)

            if save:
                save_path = self.save_dir + "/" + self.config.name
                print "saving model to {0}".format(save_path)
                saver.save(sess, save_path)
                print "saved model"

现在是我用来恢复模型的代码

        self.vocab = Vocab.load(path = restore_path + "/" + "vocab.pkl")
        self._model = create_model(len(self.vocab) + 1, config)
        self._sess = tf.Session(config=tf.ConfigProto(allow_soft_placement=True))

        saver = tf.train.Saver()
        restore_model(restore_path, self._sess, saver)

def restore_model(path, session, saver):
    saver.restore(session, tf.train.latest_checkpoint(path))

如果我使用inspect_checkpoint脚本，则输出结果为：

 python /usr/local/lib/python2.7/dist-packages/tensorflow/python/tools/inspect_checkpoint.py --file_name=models/best_reddit/language_model
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcublas.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcudnn.so.5 locally
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcufft.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcurand.so.8.0 locally
Embedding (DT_FLOAT) [19967,250]
Embedding/Adam (DT_FLOAT) [19967,250]
Embedding/Adam_1 (DT_FLOAT) [19967,250]
Projection/B (DT_FLOAT) [19967]
Projection/B/Adam (DT_FLOAT) [19967]
Projection/B/Adam_1 (DT_FLOAT) [19967]
Projection/U (DT_FLOAT) [250,19967]
Projection/U/Adam (DT_FLOAT) [250,19967]
Projection/U/Adam_1 (DT_FLOAT) [250,19967]
RNN/multi_rnn_cell/cell_0/gru_cell/candidate/biases (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_0/gru_cell/candidate/biases/Adam (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_0/gru_cell/candidate/biases/Adam_1 (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_0/gru_cell/candidate/weights (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_0/gru_cell/candidate/weights/Adam (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_0/gru_cell/candidate/weights/Adam_1 (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_0/gru_cell/gates/biases (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_0/gru_cell/gates/biases/Adam (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_0/gru_cell/gates/biases/Adam_1 (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_0/gru_cell/gates/weights (DT_FLOAT) [500,500]
RNN/multi_rnn_cell/cell_0/gru_cell/gates/weights/Adam (DT_FLOAT) [500,500]
RNN/multi_rnn_cell/cell_0/gru_cell/gates/weights/Adam_1 (DT_FLOAT) [500,500]
RNN/multi_rnn_cell/cell_1/gru_cell/candidate/biases (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_1/gru_cell/candidate/biases/Adam (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_1/gru_cell/candidate/biases/Adam_1 (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_1/gru_cell/candidate/weights (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_1/gru_cell/candidate/weights/Adam (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_1/gru_cell/candidate/weights/Adam_1 (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_1/gru_cell/gates/biases (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_1/gru_cell/gates/biases/Adam (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_1/gru_cell/gates/biases/Adam_1 (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_1/gru_cell/gates/weights (DT_FLOAT) [500,500]
RNN/multi_rnn_cell/cell_1/gru_cell/gates/weights/Adam (DT_FLOAT) [500,500]
RNN/multi_rnn_cell/cell_1/gru_cell/gates/weights/Adam_1 (DT_FLOAT) [500,500]
RNN/multi_rnn_cell/cell_2/gru_cell/candidate/biases (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_2/gru_cell/candidate/biases/Adam (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_2/gru_cell/candidate/biases/Adam_1 (DT_FLOAT) [250]
RNN/multi_rnn_cell/cell_2/gru_cell/candidate/weights (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_2/gru_cell/candidate/weights/Adam (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_2/gru_cell/candidate/weights/Adam_1 (DT_FLOAT) [500,250]
RNN/multi_rnn_cell/cell_2/gru_cell/gates/biases (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_2/gru_cell/gates/biases/Adam (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_2/gru_cell/gates/biases/Adam_1 (DT_FLOAT) [500]
RNN/multi_rnn_cell/cell_2/gru_cell/gates/weights (DT_FLOAT) [500,500]
RNN/multi_rnn_cell/cell_2/gru_cell/gates/weights/Adam (DT_FLOAT) [500,500]
RNN/multi_rnn_cell/cell_2/gru_cell/gates/weights/Adam_1 (DT_FLOAT) [500,500]
beta1_power (DT_FLOAT) []
beta2_power (DT_FLOAT) []

GPU训练模型的密钥在cpu上找不到

0 个答案: